Upon further investigation, I learned that the Post's sentence not only contradicts my theory, but also the facts! The sentence is incorrect, and taken out of context.
The Post's is (attempting to) citing the BEA numbers for Q4 released today:
(a) Actually, the -5.9% refers to the combined category of non-residential stuctures and equipment. Any time you combine those two categories, the latter will dominate the flow and the former will dominate the stock (ie, structures have much slower depreciation). So all you learn from the -5.9% is that equipment investment is down.
(b) If you look at real investment in non-residential STRUCTURES, that fell 1.5%.
(c) The -1.5% hardly changed since the BEA's last estimate Jan 30. Thus, while I was surprised at many entries in the report, commericial construction was not one of them. (in general, the quarterly GDP figures on construction do not contain surprises, because we have timely monthly releases. Quarterly equipment investment is another story).
(d) The index for real investment in non-residential construction was 109.567. The prior quarter was 111.257. The same quarter one year ago (start of the recession) was 102.076.
(e) Thus, Q4 was a small pull back (-1.5%) in strong uptrend since 2006. In this regard, commercial construction is nothing like residential construction, or consumption.
(f) Even with the BEA's revisions, investment in general in this recession is typical of the previous 3 recessions. Non-residential building is much stronger than we have seen in previous recession (see the chart below).
[It is possible that the Washington Post meant that non-residential building would shrink -5.9% if the 4th quarter repeated itself three more times. By coincidence, -0.059 is both equal to (109.56/111.257)^4 - 1 and the quarterly reduction in nonresidential investment (equipment and structures combined). In any case, the Washington Post does not know how to read the BEA report. The only question is what theory properly describes their misreading.]
[added: you will not see the error any more because the Washington Post rewrote the article over night]