@Reg
I’m sure I’m not alone in saying my eyes glazed over. What I did was ask ChatGPT this: Recently the President basically took over the agency that issues monthly job statistics, presumably so he can meddle with it and provide only favorable data with little regard to accuracy. So, I’m wondering if it’s possible to basically derive the same statistical conclusions using publicly available data.
The AI returned with this reply:
**********
It’s definitely possible to independently estimate job growth, unemployment, and related labor statistics from public data, but it takes some work and the results won’t be quite as precise or timely as the official Bureau of Labor Statistics (BLS) reports—especially if the government stops publishing them transparently.
Here’s the big picture:
We can think of the BLS jobs reports as being based on two main surveys:
Establishment Survey (CES) – surveys ~122,000 businesses and agencies about employment, hours, and earnings.
Household Survey (CPS) – surveys ~60,000 households about employment status, unemployment, etc.
If those numbers are withheld or tampered with, here’s how to replicate the core trends from alternative sources:
1. Payroll / Establishment Data
Goal: Estimate monthly job gains/losses in specific sectors.
Sources you can use:
ADP National Employment Report – private payroll processor, monthly job growth estimates.
Homebase, UKG, and Paychex/IHS Markit – small- and medium-business payroll activity data, often published weekly.
Indeed Hiring Lab & Glassdoor – job posting trends, hiring slowdown/acceleration signals.
Federal Reserve Economic Data (FRED) – aggregates multiple public/private indicators.
2. Household Labor Status
Goal: Estimate unemployment rate, labor force participation.
Sources you can use:
Census Bureau’s Current Population Survey microdata – still public if Census keeps publishing it, can calculate unemployment yourself.
Real-time population estimates from Census + private polling data from Gallup, Morning Consult, YouGov, etc., that ask about current employment status.
Household Pulse Survey (Census) – not as detailed, but can capture big shifts in employment.
3. High-Frequency Indicators
These give early signs before the monthly jobs report:
Initial Unemployment Claims – weekly, from the Department of Labor (difficult to fake because states submit it individually).
Job Openings & Labor Turnover Survey (JOLTS) – if compromised, private equivalents exist via LinkedIn or Indeed.
Transportation metrics – freight shipments, fuel consumption (proxy for industrial activity).
Small business optimism surveys – NFIB, regional Fed surveys.
4. How to Put It Together
Blend multiple indicators in a statistical model (e.g., regression or nowcasting approach) to estimate job gains/losses.
Compare to historical correlations – for example, ADP payroll growth has a fairly stable relationship with CES growth over time, though not perfect month to month.
Track revisions – often the trend matters more than the initial number.
Limitations
Private datasets don’t perfectly match the BLS methodology, so monthly noise will be higher.
Sectoral breakdowns (e.g., “manufacturing jobs” or “government jobs”) will be harder to replicate without the official survey.
Seasonal adjustment—something the BLS does meticulously—requires a good historical dataset and care to avoid misleading spikes/drops.
If you want, I can map out a step-by-step “shadow BLS” method that uses freely available data sources and some simple statistical modeling so you could track the labor market even if official releases become unreliable. That would give you a jobs estimate within a reasonable margin of error.
Do you want me to go ahead and outline that full method?
**********
If you want it to do that, just go ahead and reproduce my query.
Unseen 3 months, 3 weeks ago.
