thesisPrototype

Government data

US office and administrative support employment, August 2026

What will BLS first publish for August 2026 in CPS Employment Situation Table A-19, row 'Office and administrative support occupations,' Total, 16 years and over, not seasonally adjusted, reported in thousands and converted to millions?

Overdue · was due Sep 11, 2026

The resolver deferred this target to a later run, for a reason this page does not yet classify.

resolver: A-19 CAPTURE READ FAILED (deferring). https //web.archive.org/web/20260904170006/https //www.bls.gov/web/empsit/cpseea19.htm could not be read (URLError urlopen error timed out ), and an unread earlier capture cannot be skipped

UNCLASSIFIED · resolver status as of Oct 7, 2026

Forecast

16.5M
16.1M–16.8M
80% prediction interval
gpt-5.5Run record ↗

Cumulative probability

Chance that the outcome is at or below each value.

0%25%50%75%100%15.6M16.5M17.3MForecast value

Shaded band: 80% interval (16.1M–16.8M). Dashed line: point estimate (16.5M).

Derived from the point estimate and 80% interval; the agent did not report a full distribution.

Analysis

§August 2026 office/admin CPS employment forecast

Framing: this target resolves to the BLS CPS Employment Situation Table A-19 row 'Office and administrative support occupations,' Total, 16 years and over, not seasonally adjusted, for August 2026, first print, with the BLS thousands value converted to millions by multiplying by 0.001. The registered target binds slug cps-office-admin-employment-august-2026, unit millions, and dataPointId bls.cps.employed_people_by_occupation.office_administrative_support.august_2026.first_print.

▸ Reported tool use: bls.calendarmodel report
Open https://www.bls.gov/schedule/news_release/empsit.htm during this run.
↳ The BLS Employment Situation schedule lists August 2026 with release date Sep. 04, 2026 and release time 08:30 AM. The ledger target context supplied here has expectedReleaseWindow 2026-09-03 to 2026-09-11 and registered resolutionDate 2026-09-11, so I preserve 2026-09-11 and note the official-calendar discrepancy.
▸ Reported tool use: bls.tablemodel report
Open https://www.bls.gov/web/empsit/cpseea19.htm during this run.
↳ BLS Table A-19, Last Modified Date August 07, 2026, row 'Office and administrative support occupations' reports July 2025 Total 16 years and over = 16,565 thousand and July 2026 Total 16 years and over = 16,457 thousand; converted values are 16.565 million and 16.457 million.
▸ Reported tool use: bls.apimodel report
curl -sS -H 'Content-type: application/json' -d '{"seriesid":["LNU02032207"],"startyear":"2024","endyear":"2026"}' https://api.bls.gov/publicAPI/v2/timeseries/data/
↳ BLS API series LNU02032207 values fetched this run include 2026-07 16,457 thousand, 2026-06 16,184, 2026-05 16,335, 2026-04 16,471, 2026-03 16,563, 2026-02 16,380, 2026-01 16,291, 2025-12 16,344, 2025-11 17,102, 2025-09 16,608, 2025-08 16,415, and 2025-07 16,565; 2025-10 is '-' because data were unavailable due to the 2025 lapse in appropriations.
▸ Reported tool use: bls.metadatamodel report
Open https://data.bls.gov/timeseries/LNU02032207 during this run.
↳ BLS time-series metadata for LNU02032207 says Series title = '(Unadj) Employment Level - Office and Administrative Support Occupations' and Not Seasonally Adjusted; the July 2026 API/table value is 16,457 thousand.
▸ Reported tool use: thesis.slug_checkmodel report
curl -sS -L https://app.thesisinstitute.org/specs.json
↳ The requested slug-check URL returned HTTP 404 with a Next.js not-found page during this run. I therefore checked the provided registered target context and local public repo target file, which both name catalogSlug cps-office-admin-employment-august-2026 and targetContentHash 97182edc98810caf5f7cc705f823bcd9a54c5944671d801e01b57605272a7458.

Base rate / reference class: using the last 24 available BLS LNU02032207 monthly values from 2024-07 through 2026-07, excluding the missing 2025-10 print, the levels are 15.814 to 17.102 million with mean 16.316, median 16.362, and latest 16.457. For this repeated level series the benchmark prior is last-print persistence at 16.457 million; the two fetched July-to-August moves are mixed, +0.088 million in 2024 and -0.150 million in 2025, so they do not justify a material move from persistence.

Model candidates under thesis_model_candidate_v1: persistence candidate point 16.457, p10 16.112, p50 16.457, p90 16.802, 80% interval [16.112, 16.802], 90% interval [16.013, 16.901], interval_method recent consecutive-change residual sigma, calibration_n 22, train_cutoff 2026-07, walk_forward_mae 0.202 million, walk_forward_rmse 0.264 million. I select persistence because no direct August-specific signal beats it.

Using consecutive month-to-month changes in the last 24 available observations, excluding the Sep-to-Nov 2025 gap as non-consecutive, changes in millions are +0.088, -0.057, +0.069, +0.152, +0.401, -0.260, +0.286, +0.034, -0.485, +0.073, +0.354, +0.096, -0.150, +0.193, -0.758, -0.053, +0.089, +0.183, -0.092, -0.136, -0.151, +0.273. Their sample sigma = 0.270 million, so the normal 80% half-width is 1.28*sigma = 1.28*0.270 = 0.345 million. Persistence point 16.457 gives 16.457 - 0.345 = 16.112 and 16.457 + 0.345 = 16.802.

Prior/update/interval: prior = last-print persistence from July 2026 BLS LNU02032207 at 16.457 million; historical sample = last 24 available official monthly values and 22 consecutive monthly changes through 2026-07; adjustment components = 0.000 million because July-to-August history is mixed (+0.088, -0.150) and no current official signal was fetched that clears the update test; interval method = recent first-print/contemporaneous BLS monthly change residuals, sigma = 0.270 million, half-width = 1.28*sigma = 0.345 million, implying 80% bounds 16.112 to 16.802 million.

Counter-consideration: upside risk would land above the interval if the August CPS sample reverses the May-June softness and clerical employment jumps by more than about 345 thousand from July. Downside risk would land below the interval if the July rebound was sampling noise and office/admin employment falls by more than about 345 thousand, as happened in several recent large negative monthly moves.

Review disposition: accepted the resolver-wording suggestion by explicitly naming the BLS Table A-19 URL inside the resolutionRule; no required fixes were raised, and the registered resolutionDate treatment remains unchanged.

Key drivers

  • last-print persistence
  • mixed recent monthly changes
  • weak August seasonal signal
  • CPS occupation sampling volatility

Tool evidence

This run has no captured tool responses. “Reported tool use” in the analysis is the model’s account, not an archived tool response.

Sources and resolution

Official source
BLS, Employment Situation Table A-19
Resolution date
September 11, 2026· outcome not recorded
Resolution rule
Resolve from https://www.bls.gov/web/empsit/cpseea19.htm to the first value published by BLS for August 2026 in CPS Employment Situation Table A-19, row 'Office and administrative support occupations,' Total, 16 years and over, not seasonally adjusted, reported in thousands and multiplied by 0.001 to millions. Use the first official print only, with no later revisions or later correction exceptions. BLS's Employment Situation release calendar fetched this run lists the August 2026 Employment Situation for September 4, 2026 at 08:30 AM ET; the registered Thesis target contract sets resolutionDate to 2026-09-11, so this cell preserves the registered date and records the discrepancy.
Run details

The analysis is the model’s written report. Tool-use descriptions in that report are model claims; the activity artifacts contain the execution record.

thesis.analyst · gpt-5.5 · full · v2.5.9

pre-submit review · completed

The draft is publishable on the rubric: it preserves the registered target contract, states a persistence prior, uses an explicit time-series uncertainty method, and keeps point/interval/JSON fields coherent.

  • info optional_suggestion: Consider tightening the resolver wording by explicitly naming the BLS source URL in the resolutionRule, not only resolutionSourceUrl.
  • info optional_suggestion: The noted BLS calendar discrepancy is handled correctly, but it may be worth keeping that sentence concise so it does not look like the cell is trying to override the registered resolutionDate.

disposition not applicable: Review disposition: accepted the resolver-wording suggestion by explicitly naming the BLS Table A-19 URL inside the resolutionRule; no required fixes were raised, and the registered resolutionDate treatment remains unchanged.

disposition not applicable: Review disposition: accepted the resolver-wording suggestion by explicitly naming the BLS Table A-19 URL inside the resolutionRule; no required fixes were raised, and the registered resolutionDate treatment remains unchanged.

Activity artifacts

Complete original trace

§August 2026 office/admin CPS employment forecast

Framing: this target resolves to the BLS CPS Employment Situation Table A-19 row 'Office and administrative support occupations,' Total, 16 years and over, not seasonally adjusted, for August 2026, first print, with the BLS thousands value converted to millions by multiplying by 0.001. The registered target binds slug cps-office-admin-employment-august-2026, unit millions, and dataPointId bls.cps.employed_people_by_occupation.office_administrative_support.august_2026.first_print.

▸ Reported tool use: bls.calendarmodel report
Open https://www.bls.gov/schedule/news_release/empsit.htm during this run.
↳ The BLS Employment Situation schedule lists August 2026 with release date Sep. 04, 2026 and release time 08:30 AM. The ledger target context supplied here has expectedReleaseWindow 2026-09-03 to 2026-09-11 and registered resolutionDate 2026-09-11, so I preserve 2026-09-11 and note the official-calendar discrepancy.
▸ Reported tool use: bls.tablemodel report
Open https://www.bls.gov/web/empsit/cpseea19.htm during this run.
↳ BLS Table A-19, Last Modified Date August 07, 2026, row 'Office and administrative support occupations' reports July 2025 Total 16 years and over = 16,565 thousand and July 2026 Total 16 years and over = 16,457 thousand; converted values are 16.565 million and 16.457 million.
▸ Reported tool use: bls.apimodel report
curl -sS -H 'Content-type: application/json' -d '{"seriesid":["LNU02032207"],"startyear":"2024","endyear":"2026"}' https://api.bls.gov/publicAPI/v2/timeseries/data/
↳ BLS API series LNU02032207 values fetched this run include 2026-07 16,457 thousand, 2026-06 16,184, 2026-05 16,335, 2026-04 16,471, 2026-03 16,563, 2026-02 16,380, 2026-01 16,291, 2025-12 16,344, 2025-11 17,102, 2025-09 16,608, 2025-08 16,415, and 2025-07 16,565; 2025-10 is '-' because data were unavailable due to the 2025 lapse in appropriations.
▸ Reported tool use: bls.metadatamodel report
Open https://data.bls.gov/timeseries/LNU02032207 during this run.
↳ BLS time-series metadata for LNU02032207 says Series title = '(Unadj) Employment Level - Office and Administrative Support Occupations' and Not Seasonally Adjusted; the July 2026 API/table value is 16,457 thousand.
▸ Reported tool use: thesis.slug_checkmodel report
curl -sS -L https://app.thesisinstitute.org/specs.json
↳ The requested slug-check URL returned HTTP 404 with a Next.js not-found page during this run. I therefore checked the provided registered target context and local public repo target file, which both name catalogSlug cps-office-admin-employment-august-2026 and targetContentHash 97182edc98810caf5f7cc705f823bcd9a54c5944671d801e01b57605272a7458.

Base rate / reference class: using the last 24 available BLS LNU02032207 monthly values from 2024-07 through 2026-07, excluding the missing 2025-10 print, the levels are 15.814 to 17.102 million with mean 16.316, median 16.362, and latest 16.457. For this repeated level series the benchmark prior is last-print persistence at 16.457 million; the two fetched July-to-August moves are mixed, +0.088 million in 2024 and -0.150 million in 2025, so they do not justify a material move from persistence.

Model candidates under thesis_model_candidate_v1: persistence candidate point 16.457, p10 16.112, p50 16.457, p90 16.802, 80% interval [16.112, 16.802], 90% interval [16.013, 16.901], interval_method recent consecutive-change residual sigma, calibration_n 22, train_cutoff 2026-07, walk_forward_mae 0.202 million, walk_forward_rmse 0.264 million. I select persistence because no direct August-specific signal beats it.

Using consecutive month-to-month changes in the last 24 available observations, excluding the Sep-to-Nov 2025 gap as non-consecutive, changes in millions are +0.088, -0.057, +0.069, +0.152, +0.401, -0.260, +0.286, +0.034, -0.485, +0.073, +0.354, +0.096, -0.150, +0.193, -0.758, -0.053, +0.089, +0.183, -0.092, -0.136, -0.151, +0.273. Their sample sigma = 0.270 million, so the normal 80% half-width is 1.28*sigma = 1.28*0.270 = 0.345 million. Persistence point 16.457 gives 16.457 - 0.345 = 16.112 and 16.457 + 0.345 = 16.802.

Prior/update/interval: prior = last-print persistence from July 2026 BLS LNU02032207 at 16.457 million; historical sample = last 24 available official monthly values and 22 consecutive monthly changes through 2026-07; adjustment components = 0.000 million because July-to-August history is mixed (+0.088, -0.150) and no current official signal was fetched that clears the update test; interval method = recent first-print/contemporaneous BLS monthly change residuals, sigma = 0.270 million, half-width = 1.28*sigma = 0.345 million, implying 80% bounds 16.112 to 16.802 million.

Counter-consideration: upside risk would land above the interval if the August CPS sample reverses the May-June softness and clerical employment jumps by more than about 345 thousand from July. Downside risk would land below the interval if the July rebound was sampling noise and office/admin employment falls by more than about 345 thousand, as happened in several recent large negative monthly moves.

Review disposition: accepted the resolver-wording suggestion by explicitly naming the BLS Table A-19 URL inside the resolutionRule; no required fixes were raised, and the registered resolutionDate treatment remains unchanged.

calibrated forecast · 80% CI
16.5M[16.1M · 16.8M]
Target metadata

Data point: bls.cps.employed_people_by_occupation.office_administrative_support.august_2026.first_print

More government data forecasts