From the BLS stats in question (https://www.bls.gov/news.release/empsit.nr0.htm), the total "civilian noninstitutional population" (i.e. labor force) is 260 million. Of the 244 million who are ages 20+ and broken out by gender, there are 117.4M men and 126.0M women.
Why the difference?
- Disabled people are still included for the labor force cohort unless they're hospitalized/institutionalized, so the gender-gap impact is relatively low.
- Active-duty military service covers 1.3M adults, 90% male.
- Incarceration covers 2.3M adults, 90% male.
- The "labor force" has no upper age boundary, so the 20+ cohort has about 6 million more women. (1.5M people are in nursing homes and not counted, but the gender split is fairly even.)
That's 6 million more women total, and 3M more men than women excluded from the pool, against a laborforce gap of ~9 million. Looks pretty complete for the denominator. But the raw number of total nonfarm payrolls listed is 76.2 million per gender. That would be 66.9% male LFPR, and 62.3% female LFPR, which we know isn't right.
The BLS report instead has a male 20+ LFPR of 84M (71.5%) and female 20+ of 74.5M (59.2%), which is that 12% gap. What gives?
- Farm labor is part of the LFPR. That data is extremely messy, but there are ~3M farm workers who appear to be ~75% male.
- Multiple jobs count as multiple payrolls, but one employment. This adds 6-7M non-employment payrolls per gender to help explain why the employment counts don't match payrolls, and up to 0.5M more female payrolls than male. (Although I can't tell what percentage are nonfarm payrolls.)
That correction gets us to 70.7% male LFPR and 60.7% female LFPR, which is much closer to correct. Correct for the 16-19 numbers at it gets even closer. Looks good!
...except that apparently nonprofit jobs aren't included in nonfarm payrolls either. And that's 12M people, ~73% female. Which completely blows out the stats, so badly that I can't find any way to fix it. All I can say is either seasonal adjustment, or major differences in raw population counts between the methods (e.g. response rate gaps). Does anyone else know what's up?
From the BLS stats in question (https://www.bls.gov/news.release/empsit.nr0.htm), the total "civilian noninstitutional population" (i.e. labor force) is 260 million. Of the 244 million who are ages 20+ and broken out by gender, there are 117.4M men and 126.0M women.
Why the difference?
- Disabled people are still included for the labor force cohort unless they're hospitalized/institutionalized, so the gender-gap impact is relatively low.
- Active-duty military service covers 1.3M adults, 90% male.
- Incarceration covers 2.3M adults, 90% male.
- The "labor force" has no upper age boundary, so the 20+ cohort has about 6 million more women. (1.5M people are in nursing homes and not counted, but the gender split is fairly even.)
That's 6 million more women total, and 3M more men than women excluded from the pool, against a laborforce gap of ~9 million. Looks pretty complete for the denominator. But the raw number of total nonfarm payrolls listed is 76.2 million per gender. That would be 66.9% male LFPR, and 62.3% female LFPR, which we know isn't right.
The BLS report instead has a male 20+ LFPR of 84M (71.5%) and female 20+ of 74.5M (59.2%), which is that 12% gap. What gives?
- Farm labor is part of the LFPR. That data is extremely messy, but there are ~3M farm workers who appear to be ~75% male.
- "Proprietors", or unincorporated self-employed people, aren't counted. That's 9.5M people, roughly 68% male.
- Multiple jobs count as multiple payrolls, but one employment. This adds 6-7M non-employment payrolls per gender to help explain why the employment counts don't match payrolls, and up to 0.5M more female payrolls than male. (Although I can't tell what percentage are nonfarm payrolls.)
That correction gets us to 70.7% male LFPR and 60.7% female LFPR, which is much closer to correct. Correct for the 16-19 numbers at it gets even closer. Looks good!
...except that apparently nonprofit jobs aren't included in nonfarm payrolls either. And that's 12M people, ~73% female. Which completely blows out the stats, so badly that I can't find any way to fix it. All I can say is either seasonal adjustment, or major differences in raw population counts between the methods (e.g. response rate gaps). Does anyone else know what's up?