Okay, strap in for yet another tirade on why you should (for the most part) pay very little attention to the monthly Non Farm Payroll data, also know as the Employment Situation. Its hype to usefulness ratio is abysmally low.
Long time readers know this is a peeve of mine (See list of prior articles below). For newer readers, here is the reasoning behind my disdain.
The first has to do with the idea of modeling the real word via assembled data. What you end up with is some form of artificial creation that bears only a passing resemblance to the complexity of the real world. That would not be such a problem, if only you humans remembered what the great professor George E. P. Box advised us: “Essentially, all models are wrong, but some are useful.”
The usefulness of a model that is wrong is that we can into what us useful and ignore what is wrong. And we know from each month’s revisions that the initial read is off, often by a substantial amount. It’s a noisy series, subject to many errors and subsequent corrections.