Daniel Zhao

"Preliminary research, which is not comprehensive and is subject to updates in QCEW data, indicates that the primary
contributors to the overestimation of employment growth are likely the result of two sources—response
error and nonresponse error." (4/9)

A nice thread from Guy on JP's comments yesterday on NFP growth being overstated.

To which I will add a caveat - the birth/death model is not the principal driver of the wedge between CES and QCEW these last two years.

(Warning, statistical esoterica ahead) 🧵1/9

Employee sentiment dropped further in the November Glassdoor Employee Confidence Index. 44.6% of employees reported a positive business outlook for their employers in November, down from 45.9% in October. Despite the shutdown ending, worker anxiety remains high.

www.glassdoor.com/blog/glassdo...

Ugh thank you, will delete and fix

Initial unemployment insurance claims rose to 236,000 last week, a sharp rebound after an unusual drop the prev week (perhaps due to some unadj. Thanksgiving seasonality).

Continuing claims similarly fell sharply, but may be facing the same issue (continuing claims are lagged an extra wk)

Upcoming Changes to the Establishment Survey Birth-Death Model Effective with the release of January 2026 data, the establishment survey will change the birth-death model by incorporating current sample information each month. The change follows the same methodology applied to the April through October 2024 forecasts during the 2024 post-benchmark period (see question 9 in the CES Birth-Death Model Frequently Asked Questions page at www.bls.gov/web/empsit/cesbdqa.htm).

screenshot question 9 from BLS CES birth-death model Q&A: https://www.bls.gov/web/empsit/cesbdqa.htm relevant quotes pasted below: "BLS releases the Quarterly Census of Employment and Wages (QCEW) monthly employment data on a quarterly basis 6 to 9 months after a given collection month. During current month birth-death processing, the forecasted portion of the birth-death model uses inputs derived from comparing the following the QCEW-based employment data including business openings and closings and the same QCEW data excluding the birth units and imputing monthly values to the death units. The difference between these two QCEW-based monthly values is the net of births and deaths. ... These ... can produce accurate forecasts when business openings and closings for current months follow predictable seasonal patterns. However, sharp trend changes or extreme events like hurricanes and pandemics can result in large forecast errors. For the 2024 benchmark release, BLS included a regression component to the time series model. The model exploits the linear relationship between the natural log of the sample-based ratio of employment change, also known as the weighted-link-relative (WLR), and the actual birth-death input values derived from lagged QCEW data described above. ... Although the QCEW data are not available until 6 to 9 months after a given month's reference period, the sample-based ratio of employment changes is available as soon as the monthly reference period is over. Thus, the relationship between the WLR and the QCEW-based birth-death inputs can be modeled using historical data and predicted for more recent months using the sample-based ratio. This adjustment makes the total net birth-death forecasts more accurate during both relatively stable time periods and during more volatile shifts in trend. ... the expected benefits in accuracy for the post-benchmark period were substantial and warranted using this approach during that period."

Missed this at the time but the Sep jobs report announced changes to the net birth-death model effective with the Jan 2026 report release that seems promising. Should improve the accuracy of the model and hopefully reduce size of benchmark revisions

Q9: www.bls.gov/web/empsit/cesbdqa.htm

afaik, JOLTS still samples fed govt establishments so there's probably pretty wide error bars esp for something as spiky as the deferred resignation program. otoh, CES has almost universal coverage of federal payrolls so next week's jobs report will likely more accurately capture the change

not sure i follow, that table screenshot from the report does show a sharp rise in federal worker separations in Sep

bsky.app/profile/dani...

The current standard, where latinos are supposed to indicate that their race is white and ethnicity latino is confusing and is messing up demographic data (‘other race’ is exploding as a category because of this) and I was glad that this was going to get fixed. Guess now change is on pause.

A nice piece on algorithmic pricing worth your time.

A question I got frequently when I led BLS was why the agency went into stores and didn’t just collect prices online. My answer, “prices online are often different than in the store, so we do both” often surprised people.

(absent any future moves) it'll likely be the most noticeable spike in the data. an ongoing trickle of cuts on the order of hundreds/thousands aren't large enough to move the data. slower hiring is having a big impact too but is spread out rather than spiky

fred.stlouisfed.org/series/JTU91...

talks? do you mean the data impacts from the cuts, deferred resignation program, etc.? i think we'll see more clarity on that in the next jobs report

Overall, imo, trend in headline data in today's report looks a little worse than actual (volatility in accommodation & food services may be driving layoff 🔼, quit 🔽 in Oct), but hires are still both sluggish & trending down. Job market not spiraling out of control but still ongoing cooling.

8/8

Federal quits spiked in Sep to a record 47,000, in line with the effective date of the deferred resignation program. Similarly, other separations (includes retirements) for federal workers spiked to 27,000.

Shows some of the impact of federal job cuts though likely doesn't capture them all

7/

Note of caution about the 1-month drop in quits/jump in layoffs: much of that is driven by accommodation & food services, where data has been unusual lately. Quits in the sector surged in the summer & layoffs have more than doubled in 2 months. Could point to data issues.

6/

Surprisingly, job openings actually increased over the last 2 months rising to 7,670,000 in Oct from 7,658,000 in Sep and 7,227,000 in Aug.

Openings are usually volatile but seem to have been moving sideways in JOLTS data since mid-2024, even if the hires rate has slowed over the same period.

5/

Layoffs rose to 1,854,000 in Oct, highest since Jan 2023. On a rate basis, layoffs rose to 1.2%, highest since Sep 2024.

Layoffs have been creeping upward on average over the year though remain in a similar range to pre-pandemic (both in level and rate).

4/

The quits rate tumbled to 1.8% in Oct, lowest since May 2020. Like hiring, quits have also been sluggish, similar to 2014 levels.

Diminished quits indicate workers are not finding other opportunities on the open market/are not confident in their ability.

3/

The hires rate was 3.2% in Oct, down from 3.4% in Sep, but unchanged from Aug.

3.2% ties the lowest post-Covid level and is comparable to sluggish hiring levels in 2013 coming out of the 2008 recession

2/

#JOLTS is back from shutdown with Oct report today including Sep + Oct data:

-Hires rate 3.2%, down from 3.4% in Sep flat from Aug
-Quits rate down to 1.8% from 2.0% in Sep, 1.9% in Aug
-Openings up to 7.67m from 7.66m in Sep, 7.23m in Aug
-Layoffs jump to 1.85m from 1.78m in Sep, 1.73 in Aug

1/

Congestion pricing improved air quality in NYC and suburbs | Cornell Chronicle Cornell researchers tallied the environmental benefits of New York City’s congestion pricing program and found air pollution dropped by 22% in Manhattan, with additional declines across the city’s fiv...

In the first six months of congestion pricing, air pollution within Manhattan's toll zone dropped 25%.

Pollution fell across the five boroughs and in the suburbs, too. A clean air win all around.

Bank of America payroll estimate has been trending even worse than ADP recently. Latest data from Nov at +0.2% payroll growth year-over-year.

*They only produce a 3-month average of year-over-year % change

institute.bankofamerica.com/economic-ins...

The only value in the database for JOLTS that won’t be present will be the unemployed persons per job opening (JTS000000000000000UOR). It will be represented by the presence of a record but a missing value in the variable (noted with a hyphen). In our revision tables, the missing second closing August data and missing first closing September data will be noted with a #N/A in the appropriate cell.

Advance insight to help you code around a new, stupid issue.

How will BLS represent missing data values resulting from the govt shutdown in the #JOLTS data release tomorrow?
#NumbersDay
cc: @bencasselman.bsky.social @nickbunker.bsky.social @jc-econ.bsky.social

In looking at explanations for ^, I came across this interesting tidbit:

Even though private edu & health services have accounted for ~70% of private job gains over the last yr, wage gains there have been sluggish

Avg weekly earnings are up 1.5% YoY in the sector vs. +3.8% for all private

The gender pay gap widened in Q3 2025 to the widest since Q1 2020. The ratio of full-time working women's earnings to men's fell to 80.9% in Q3, a continued deterioration from a record 84.6% in Q2 2023.

Source: BLS, www.bls.gov/news.release...

Merry Christmas from the BEA 🎄

Antitypical's comment on "[Schatz] [Yesterday's Vikings game] would be the worst Jets offensive game ever if it had been played by the Jets, over 2023 Week 15 and 2019 Week 5. THE JETS." Explore this conversation and more from the nfl community

fun suggestion from reddit:

> see how many franchises this would be the worst offensive game for ... Can you imagine a post being like "yesterday's Vikings game would be the worst offensive performance in history for 27/32 franchises, including the New York Jets"

www.reddit.com/r/nfl/commen...

Community college enrollment has been declining since 2009/10. The optimal policy response to this depends on the root of this decline.

I'm thrilled @nber.org today released my working paper with Harvard PhD Joe Winkelmann titled:

"Labor Market Strength and Declining Community College Enrollment"

Screenshot of Venmo post-payment offer for DraftKings

Paid my rent today and got a sports gambling ad afterwards 🤔

BofA: We think tighter immigration policies will reduce the breakeven pace of employment growth to about 20k