There’s plenty of attention being paid today to the fact that employment as measured by the BLS household survey really surged, while the payroll survey just offered moderately good news. The household survey presumably has some real epistemic value, but I think it’s important to try to not opportunistically switch from one dataset to the other. But however you like this, it’s worth noting that the household survey always finds a lot more jobs than the payroll survey.
You’ll see above a chart of the gap in terms of thousands of jobs. We’re talking a gap that ranges in scale from hundreds of thousands of jobs to over a million jobs. It’s a very large gap. And unfortunately for analysts, it’s not a stable gap either. It looks a little bit like labor market booms—as in the mid-sixties or the late-nineties—cause the gap to close, but it’s quite noisy on a month-to-month basis and seems to have other mysterious sources of variation.
If you recall the data disputes of the mid-aughts when the gap was rising, we used to have a lot of accusations and counter-accusations flying around. Sometimes liberals would suggest that the Bush administration was deliberately cooking the books on the household survey (but for some reason forgetting to do so on the payroll survey) and that’s why the gap was rising. And conservatives sometimes argued that we were seeing a surge in home-based businesses (eBay often got a mention) that the payroll survey wasn’t finding. I wonder in retrospect if undocumented workers in the construction sector weren’t playing a role here. At any rate, we never really figued it out and the discrepency never went away. Instead it stabilized at a level higher than what we saw in the nineties but lower than what we saw in the eighties.
I have genuinely no theory to offer about this, just the observation that when you hear theories offered it’s important to try to locate them in the broader context of longstanding and variable disagreement between the series.