<< Read Part 1: Really Looking For Inflation
What these unusually weak productivity estimates lean toward is, quite simply, the possibility the BLS has been overstating jobs gains for years. In early 2018, there is already the hint of just that problem in a 4.1% unemployment that doesn’t lead to any acceleration in wages and labor income. What it does suggest is that something (or several somethings) in these estimates is off somewhere.
For the unemployment rate, that already includes the participation problem in its denominator, but, again, that is not mutually exclusive of problems in the numerator (the increase in the number of payrolls). As nothing more than a rhetorical exercise utilizing nothing more than back-of-the-envelope counterfactuals (so take it in that spirit), if productivity had been more balanced and thus more consistent with how an economy actually works over the intermediate and long terms (not transitory), that would have meant by simple arithmetic either output was much higher or labor input much lower.
The Household Survey gained 1.44% per year during those same years, a lower rate than total hours worked reflecting the increase in full-time jobs as some part-time positions were converted back to the former pre-crisis status. Reducing the total gain in hours worked by more than a third (as shown above) would have lowered the increase in the Household Survey by more than 5.2 million at the end of 2017, leaving out how in every likelihood the reduction would have been more severe factoring less part-time jobs conversions.
Assuming the labor force increased by the same rate as it has (which isn’t unreasonable given that it hasn’t increased by all that much in the current estimates), that would give us an unemployment rate around 7.4% rather than 4.1%. And if growth in the labor force was overstated in somewhat the same way as suggested in the Household Survey (with the Establishment Survey far worse), then the unemployment rate would be a little lower than that, somewhere around 6.5% (give or take).
