Sign in

Aaron Sojourner

@aaronsojourner.org
32K followers 7.5K following 18K posts

Labor econ. Formerly White House CEA senior economist for labor, UMN prof, union organizer, bike courier, dishwasher. Minneapolis. Posts mine, not employer's (@upjohninstitute.bsky.social). Fellow @iza.org aaronsojourner.org Be kind. Work hard. Have fun.

PostsRepliesMedia
Aaron Sojourner @aaronsojourner.org · 15h
"moderate, monthly unconditional $ transfer alone... insufficient to make sustained & meaningful improvements in mothers' time spent in enriching activities, child-specific expenditures, or in the quality of mother–child interactions...." onlinelibrary.wiley.com/doi/abs/10.1...
160
Aaron Sojourner @aaronsojourner.org · 04/10/2026
If only the Constitution gave Congress power over the federal budget instead of the president. Ah well, nevertheless...
1017260
Aaron Sojourner @aaronsojourner.org · 02/10/2026
This opening cracks me up
Executive Summary
October 2018
Coincident with the 200th anniversary of Karl Marx’s birth, socialism is making a comeback in
American political discourse. Detailed policy proposals from self-declared socialists are
gaining support in Congress and among much of the electorate.
2351
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Like deja vu all over again trumpwhitehouse.archives.gov/wp-content/u...
030
Aaron Sojourner @aaronsojourner.org · 02/10/2026
330
Aaron Sojourner @aaronsojourner.org · 02/10/2026
I wonder what Chris Phelan's gotten up to since becoming chair of CEA? www.whitehouse.gov/wp-content/u...
5253
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Surely this chart will cheer folks up! Official inflation measures have been quite accurate relative to those based on alternative, more-detailed data. Seems inflation has actually been a bit LOWER than official estimates. @besttrousers.bsky.social
4406
Aaron Sojourner @aaronsojourner.org · 02/10/2026
yeah, those revisions are not particularly large. This document from Friends of BLS discusses why revisions occur static1.squarespace.com/static/5bedb...
Table 1: Sources of Adjustment to Monthly Estimates
Source of
adjustment Explanation 2025 Context
Sample
discrepancies
Out of sample employers can have
faster or slower growth than those in
the CES sample.
Employer participation in the CES has
fallen sharply over the past decade. That
raises the risk of modest overstatement
when hiring slows and firm behavior
becomes more uneven. In 2025,
employers who participated tended to
have higher job growth rates, even when
controlling for industry and size.
Reporting
corrections
CES respondents may systematically
misreport their employment change
compared to their QCEW reports.
Sources of CES misreporting can
include random error and erroneous
inclusion of contract or informal
workers, such as those related to
immigration status.
BLS research indicates that CES may
have immigration-related measurement
gaps in 2025. In effect, employers
reported more employment of
immigrants in the CES than they
reported in their unemployment
insurance reports. That mismatch can
widen differences between preliminary
estimates and benchmarked totals, often
in a downward direction.
Birth-death
model
Because CES surveys existing
employers, it cannot identify business
closures or new firms as they happen.
As a result, it does not fully capture
job gains or losses from openings,
closures, or delayed reports in real
time. To account for these changes,
BLS applies a statistical birth-death
model that estimates the net
employment changes resulting from
employer starts and stops each
month.
The BLS publishes QCEW estimates as
soon as they are available and recent
ones show that a downward revision is
likely to arrive in the February 2026
Employment Situation report. Although
the exact details won’t be known until
then, BLS has studied model
performance and reported that the
current model likely overstated newfirm job creation as the labor market
cooled from its post-pandemic surge.
130
Aaron Sojourner @aaronsojourner.org · 02/10/2026
the data is much lower quality, difficult to access, and noisy by occupation but construction industry has a high degree overlap with a given set of occupations so I don't think it would change much like almost all carpenters work in construction
020
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Agree with the 2nd part. Here's the trend in the rate of construction job change over the year in percentage terms.
000
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Construction jobs are up. Lots of data centers.
3150
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Last thing. @usbls.bsky.social does amazing work to create timely, accurate info about America's working families, a huge public good They're there for us & we need to show up for them If u r a labor economist or care abt workers & employment, follow & join Friends of BLS www.friendsofbls.org/join
0114
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Firing rate also very low
3112
Aaron Sojourner @aaronsojourner.org · 02/10/2026
For context, the hire rate is very low, like in 2011. This is from #JOLTS data, not #JobsDay
1102
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Looking at change rates in wages versus consumer prices 3-month annualized, in the most-recent data... Wages (red) up at a 2.4% rate Prices (blue) up at a 0.2% rate So real hourly wages increasing at abt 2.2% annualized rate If continues, average hour'll buy 2.2% more than year before
181
Aaron Sojourner @aaronsojourner.org · 02/10/2026
If employer demand for labor falls or more workers enter to supply, price (wage) falls Private-sector wage growth continues to decelerate. Supply not 🔼, so demand 🔽 Annualized rates all decelerated: over-month: 1.6% over-3-months: 2.3% over-year: 3.0% Bizs & retirees want🔽. Working families🔼
1122
Aaron Sojourner @aaronsojourner.org · 02/10/2026
That’s the extensive quantity (Q) margin. Intensive Q margin = average hours. Average workweek hours for private-sector steady at 34.4 hrs. Normal.
1112
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Prime age (25-54 years old) EPOP at 80.7% measures core labor market strength, omits folks on fringes of work. Up 0.3 pp from Jul-Aug. Down 0.2 pp from record-high of 80.9% in July-Sept 2024. Never higher than this during 2002-2022 (80.6% high) Constrains job growth without immigration
1112
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Both the labor force participation rate (LFPR), at 61.8%, and the employment-population ratio (EPOP; figure), at 59.2%, changed little in Sept. These show little net change since Jan.
1154
Aaron Sojourner @aaronsojourner.org · 02/10/2026
The unemployment rate edged up to 4.2% from 4.1% in recent months
1194
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Job growth may be decelerating: +41 K/mo over most-recent 12 months +66 K/mo over most-recent 6 months +51 K/mo over most-recent 3 months +29 K this past month Huge amounts of uncertainty reduces businesses’ willingness to invest in jobs
59628
Aaron Sojourner @aaronsojourner.org · 02/10/2026
How does recent growth compare to recent years? Economy added 0.5 million jobs over 12 months to Sept 2026. Slowest growth in many years outside COVID.
15019
Aaron Sojourner @aaronsojourner.org · 02/10/2026
+29K jobs gained last month (mid-Aug to mid-Sept). -60K jobs in revisions of prior 2 months. The big picture
48429
Aaron Sojourner @aaronsojourner.org · 02/10/2026
Last day of the season sale at Nelson's on Snelling: $2.50 any size
4160
Aaron Sojourner @aaronsojourner.org · 01/10/2026
Looks who's on SOM's home page! Kerwin Charles is many things and crazy is not one of them.
140
Aaron Sojourner @aaronsojourner.org · 01/10/2026
To get some sense of whether hiring on the basis of a question’s responses might lead to disparate impact, we simulate what would happen if one simply hired according to each topic score. Would there be extremely disproportionate hiring by gender, race/ethnicity, or age?
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
Machine (GPT)-human agreement in coding is similar to human-human agreement. So the topic definitions aare clear and the machine does a good job classifying documents into topics.
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
We end up defining 4 topics and giving them names and descriptions, which serve as top-down coding instructions for human RAs and LLMs.
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
It’s grounded in bog-standard, qualitative data analysis methods. Then we make informed judgement about how to define final topics out of these clusters. The method gives us a variety of tools to make these judgments in knowledge of the structure of the data.
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
We cluster the codes up to dozens of code clusters using those outputs. Each cluster ties back to specific codes, and quotes in specific documents. This gives much more interpretability, transparency, and auditability than pure embedding approaches or LDA.
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
We walk through analysis of the classroom management question in detail but do each Q similarly. We ask the LLM to extract 1 or 2 “codes” from each applicant response document. For each code, provide a name, an embodying quote, and a code description. For example,
110
Aaron Sojourner @aaronsojourner.org · 01/10/2026
What do teacher job applicants discuss when answering open-ended questions & which Qs differentiate applicants? We (w/@cedr.bsky.social) develop a new LLM-assisted approach to inductive topic modeling cuz we didn't want a black-box or top-down model. Code 4u in appendix www.iza.org/publications...
IZA Discussion Paper No. 18942
September 2026
Eliciting Information about Teacher Applicants Using Open-Ended Questions
Joseph Aguilar-Bohorquez, Dan Goldhaber, Cyrus Grout, Elton Mykerezi, Prayash Pathak, Aaron Sojourner
Employers often use open-ended questions in hiring, but usually responses aren't recorded, and knowledge about the usefulness of such responses is limited. This study investigates properties of written responses to open-ended questions collected over many years via an application tracking platform that school districts use to hire teachers. We develop a scalable process to discover topics within the corpus of responses to each question and to score documents. It adapts a traditional qualitative topic modeling approach, Inductive Thematic Analysis, to combine the strengths of researcher judgment and large language models while preserving interpretability and transparency. Questions elicit distinct information regarding a candidate’s pedagogical skills, classroom management, professional philosophy and growth mindset, relational and cultural competence, and other professional attributes. Questions regarding classroom activities and student progress generate more dispersion in responses across applicants and fewer extreme demographic selection ratios. This framework empowers researchers and practitioners to assess language-based hiring instruments transparently and to monitor potential demographic disparities.
142
Aaron Sojourner @aaronsojourner.org · 01/10/2026
Employers push up employee salaries to avoid overtime obligations. Workers end up making more. "This subset of bunched workers experienced an estimated 8.5 percent increase in base salary, while other affected workers saw comparatively modest gains from overtime pay." www.nber.org/digest/20261...
0133
Aaron Sojourner @aaronsojourner.org · 30/09/2026
Advance U.S. International Trade in Goods

09/30/2026 08:30 AM EDT
The advance international trade deficit in goods increased to $132.6 billion in August from $118.9 billion in July as imports increased more than exports.

August 2026: 11.5° % Change
July 2026: 17.7° % Change
2183
Aaron Sojourner @aaronsojourner.org · 30/09/2026
On the SEED Act's expansion of the tax deduction for educator out-of-pocket expenses from K-12 to include ECE teachers, the teeniest step in the right direction. www.yahoo.com/news/us/arti...
Experts called it a "symbolic" win more than a policy that would put significant dollars back in educators' pockets given its relatively modest amount for the over 90% of Michigan early educators research shows pay for items – diapers, books, craft materials and clothing – for kids in their classrooms with their own cash.

"It's an opportunity for (early educators) to be seen as equivalent to their K-12 peers," said Aaron Sojourner, senior economist and researcher at the W. E. Upjohn Institute for Employment Research in Kalamazoo.

"Often times they have the same credentials and degrees but are just working with a different age group ... I think the bigger part is about the feeling of disrespect and lack of recognition in the field for people working with kids prior to them arriving at those Kindergarten doors,"
051
Aaron Sojourner @aaronsojourner.org · 29/09/2026
Consider signing on. Email me if you want the links for sign on.
Good morning,

We are very concerned about recent revisions made to the Commerce Department’s Scientific Integrity Policy (DAO 216-23) which removed language explicitly banning policial interference and providing whistleblower protections. Producing objective, accurate, and relevant statistics is the hallmark of our federal statistical system and removing this language unnecessarily walks back safeguards for these data.



We have drafted a letter to Secretary Lutnick expressing our opposition to these revisions and asking that this language be restored.



You can read the letter and add your signature using the appropriate link below. We are accepting sign-ons from organizations and individuals until COB Wednesday, October 7.



Organization Sign-on Form



Individual Sign-on Form

Thank you for your consideration.The undersigned organizations and individuals, representing a broad coalition of scientific, statistical, and professional communities dedicated to the integrity of federal data, write to express our concern regarding recent updates to the U.S. Department of Commerce Scientific Integrity Policy (DAO 216-23). Specifically, we urge the Department to restore language:

1. Prohibiting political interference or inappropriate influence in the collection, analysis and dissemination of federal data.

2. Providing explicit whistle-blower protections for people who report scientific integrity concerns.

American businesses, institutions and organizations rely on objective, evidence-based data produced by the Department’s principal statistical and scientific bodies, including the U.S. Census Bureau and the Bureau of Economic Analysis (BEA). Eliminating unambiguous language designed to protect government science from political manipulation significantly compromises the institutional safeguards that sustain Americans’ trust and reliance upon our national data infrastructure.

Our collective concerns focus on two critical threats to federal data:

    Erosion of Public and Professional Trust: Federal economic and demographic indicators drive billions of dollars in public and private investments. By removing explicit firewalls against political pressure and inappropriate influence, the revised departmental policy would substantially undermine the credibility of the data which will be difficult to reclaim.
    Vulnerability of Career Scientists: Clear, enforceable rules against political interference and inappropriate influence provide necessary professional protection for federal researchers and statisticians, ensuring that the data products they compile and release to the public adhere to well-established and widely accepted scientific standards and have not been judgmentally altered or skewed with political considerations in mind. Removing this language leaves agency experts vul…
162
Aaron Sojourner @aaronsojourner.org · 29/09/2026
Hiring rate remains low, like in 2011.
041
Aaron Sojourner @aaronsojourner.org · 29/09/2026
American employees have enjoyed a long streak of record low firing rates. In our 2 decades of data pre-pandemic, the smallest share of employees fired in any month was 1.1%, happened in just 2% of months. Since Jan 2021, the share fired has been at or below 1.1% in 90% of months including Aug'26.
140
Aaron Sojourner @aaronsojourner.org · 29/09/2026
1.87 American employees quit their job for each one fired in Aug. This #LaborLeverageRatio proxies for worker bargaining power in the job. Pretty flat over last 2 years, below immediate pre-pandemic level. www.upjohn.org/labor-levera...
173
Aaron Sojourner @aaronsojourner.org · 28/09/2026
He's just trying to shift blame for the fact that his decision to start another war in the Gulf caused huge price inflation. www.investopedia.com/fuel-price-s...
1183
Aaron Sojourner @aaronsojourner.org · 27/09/2026
Tough Lynx loss to NY Liberty in playoff game 1.
041
Aaron Sojourner @aaronsojourner.org · 26/09/2026
@unidosmn.bsky.social is leading on this. Check it out. unidos-mn.org/climate-equity
165
Aaron Sojourner @aaronsojourner.org · 25/09/2026
The mistake at that time was trusting that Hamas would be peaceful & not acting on the timely warning.
100
Aaron Sojourner @aaronsojourner.org · 25/09/2026
@econofact.bsky.social for Educators provides resources for #TeachEcon #EconSky econofact.org/econofact-fo...
040
Aaron Sojourner @aaronsojourner.org · 25/09/2026
11.2 million environmentalists didn't vote in 2024 To encourage 2026 voting, @patagonia.com will reward you with a $100 credit if you get 3 others to make a vote plan www.patagonia.com/vote/
The company announced today it will launch the 11.2 Million Vote Project this Saturday, Sept. 26, which is Public Lands Day. There are 90 million acres of public land at threat across the U.S, and research from the Environmental Voter Project found that 11.2 million registered-to-vote environmentalists did not vote in the 2024 election.  

The outdoor apparel business will ask its U.S. customers and community to encourage friends and family to make a plan to vote in the upcoming midterms. Participants who get three people to create a plan to vote will earn a $100 credit to redeem on patagonia.com or in a Patagonia retail store, or they can opt for Patagonia to donate $100 to the League of Conservation Voters
23215
Aaron Sojourner @aaronsojourner.org · 25/09/2026
Tax social media ads Fund local community institutions @chrismurphyct.bsky.social's IRL Act www.murphy.senate.gov/newsroom/pre...
Sen. Chris Murphy's the In Real Life (“IRL”) Act—as part of his policy agenda aimed at record-high rates of loneliness and distrust by holding social media companies accountable and investing in third spaces. The legislation would tax digital advertising revenue of only the very largest algorithmic-based social media corporations at 50%, redirecting funds to a newly-established Social Capital Fund for local communities to support the institutions, programs, and physical spaces that bring neighbors together in common cause.
2235
Aaron Sojourner @aaronsojourner.org · 25/09/2026
Been a while since we shared an office in grad school but still great to hang out. #EconSky
0160
Aaron Sojourner @aaronsojourner.org · 24/09/2026
6abc.com/amp/post/tru...
000
Aaron Sojourner @aaronsojourner.org · 24/09/2026
6818
Aaron Sojourner @aaronsojourner.org · 23/09/2026
Co(verup)ncerning
2213