The article doesn't talk about how productivity is measured and the actual sensitivity and accuracy of the measurement(s). As others have pointed out, measuring productivity is a little hazy and probably is influenced by subjective bias (hey a recession might be coming... hmm, I feel like less work is getting done at my office).

I wonder how much bias affects the reported measurements? I doubt businesses outside of manufacturing can actually discern a 2% change in productivity, when screening for other factors, and some of them can't discern a 5% change.