thesisPrototype

Government data

Belgium consumer confidence, August 2026

National Bank of Belgium consumer confidence indicator for August 2026, seasonally adjusted, first official print

Overdue · was due Aug 21, 2026

The resolver has no route for this target's reference. The target was registered with a generic source link rather than a resolver binding.

NO_ROUTE_GENERIC_URL · resolver status as of Oct 7, 2026

Forecast

-5.0
-11.0–+1.0
80% prediction interval
gpt-5.5Run record ↗

Cumulative probability

Chance that the outcome is at or below each value.

0%25%50%75%100%-20.0-5.0+10.0Forecast value

Shaded band: 80% interval (-11.0–+1.0). Dashed line: point estimate (-5.0).

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

Analysis

§Belgium August 2026 NBB Consumer Confidence Forecast

Framing and exact resolver: the target is the National Bank of Belgium's first published August 2026 consumer confidence indicator, seasonally adjusted, in signed index points. The canonical ledger target uses country BE, unit index_points, dataPointId nbb.consumer_confidence.indicator.2026-08.first_print, and an expected release window from 2026-08-16 through 2026-08-24.

▸ Reported tool use: official.lookupmodel report
Fetched the NBB upcoming latest-figures view for consumer confidence releases.
↳ Fetched official schedule numbers: Firms and short-term business statistics, period 2026M8, Consumer confidence indicator, Publication, release date 21 Aug 2026; the same page's August calendar grid shows 2 results for 2026-08-21.
▸ Reported tool use: official.lookupmodel report
Fetched the official NBB July 2026 consumer confidence indicator release page and extracted the headline table.
↳ Fetched NBB release numbers: release date 22 Jul 2026 11:00, period 2026M7; consumer confidence indicator values were 2025 Jul -4, Aug -2, Sep -1, Oct 0, Nov 2, Dec -1, 2026 Jan 4, Feb 1, Mar -6, Apr -9, May -10, Jun -7, Jul -5.
▸ Reported tool use: official.lookupmodel report
Fetched the NBB consumer-confidence methodology page for target identity and construction.
↳ Fetched methodology numbers: the survey covers approximately 1850 households, interviews run between the 1st and 15th days of each month, the indicator is the arithmetic mean of 4 seasonally adjusted key question balances, and balances are bounded from -100 to +100.
▸ Reported tool use: local.modelmodel report
Computed thesis_model_candidate_v1-style one-step benchmarks from the fetched NBB history.
↳ Model candidate numbers from fetched history: persistence point -5.0, p10 -9.3, p50 -5.0, p90 -0.7, 80% interval [-9.3, -0.7], 90% interval [-10.5, 0.5], train cutoff 2026-07, calibration_n 12, walk-forward MAE 2.75; recent monthly changes were 2, 1, 1, 2, -3, 5, -3, -7, -3, -1, 3, 2.
▸ Reported tool use: network.checkmodel report
Checked the requested public specs slug endpoint before finalizing.
↳ Fetched https://app.thesisinstitute.org/specs.json and received HTTP 404 with an HTML body of 11289 bytes rather than a specs JSON document; the registered slug belgium-consumer-confidence-august-2026 was kept.

Base rate/reference class: using the 13 official NBB monthly first prints from 2025-07 through 2026-07, the latest print is -5.0, the 13-print mean is -2.9, the last-six mean is -6.0, the last-three mean is -7.3, and the range is -10.0 to 4.0. Persistence is the base rate because it has the lowest walk-forward MAE among the simple candidates.

Level, momentum, one-off, and policy mechanisms: the level remains negative but has improved from May -10 and June -7 to July -5; momentum is upward, but the July table also shows unemployment expectations at 14, worse than June 13, while savings improved from 19 to 23. I have no direct public August survey signal before the first print, so I do not move materially away from the persistence benchmark.

Prior/update/interval: persistence prior = -5.0 from the July 2026 first print; historical sample = NBB 2025-07 through 2026-07 first-print table; adjustment components = +0.0 for August because July's improvement and unemployment deterioration are already in the latest print and no direct August signal is fetched. For this level survey-balance series, use successive monthly changes: sigma = 3.37 from changes [2, 1, 1, 2, -3, 5, -3, -7, -3, -1, 3, 2]. The raw 80% half-width is 1.28*sigma = 4.31; widen by about 1.39x to 6.0 for small-sample calibration and conservative one-month survey outcome volatility, giving -5.0 - 6.0 = -11.0 and -5.0 + 6.0 = 1.0.

Counter-consideration and falsification: upside risk is a stronger August improvement in household saving and macro expectations, which would land above the interval if the first print is 1.1 or higher. Downside risk is a renewed deterioration in unemployment or financial-situation expectations, which would land below the interval if the first print is -11.1 or lower. Outside the interval would require a larger one-month move than most recent NBB changes.

Review disposition: accepted the interval critique by removing calendar/source-binding uncertainty as a value-interval driver and tying the 6-point half-width to small-sample calibration and conservative survey outcome volatility. Accepted the clarity suggestion by stating that the official NBB calendar says 2026-08-21. Kept the specs endpoint check because it was actually fetched during the draft run.

Key drivers

  • July first print improved to -5 but remained negative
  • persistence is the strongest simple benchmark
  • July savings and economic-expectation balances improved
  • July unemployment expectations worsened slightly
  • recent monthly change sigma is 3.37 index points

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
National Bank of Belgium consumer survey release
Resolution date
August 21, 2026· outcome not recorded
Resolution rule
Resolve to the first National Bank of Belgium published consumer confidence indicator for 2026-08, seasonally adjusted, reported in signed index points and rounded to the agency's published precision. Use the first official NBB print only; ignore later NBB.Stat revisions, corrections after the first release, FRED, and third-party mirrors. The official NBB calendar fetched in this run says the 2026M8 consumer confidence indicator publication date is 2026-08-21. The registered source URL is https://www.nbb.be/en/statistics/business-surveys/consumer-survey; during this run that URL returned an NBB 404 shell, while current official NBB latest-figures pages provided the calendar and history, so keep the forecast tied to the same registered target and source binding.
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

Draft is broadly publishable, with one interval-rationale cleanup needed before publication.

  • warning interval: The 80% interval is widened from the realized monthly-change sigma using 'source-binding/calendar uncertainty,' which is not uncertainty about the target value itself.
  • info optional_suggestion: In the final reasoning, make the release date statement explicit as 'official NBB calendar says 2026-08-21' so it is clear why resolutionDate differs from the ledger window end.
  • info optional_suggestion: Consider dropping the failed specs.json check from sourceContext unless the validation pipeline needs it, since it adds noise without supporting the forecast.

disposition accepted: Review disposition: accepted the interval critique by removing calendar/source-binding uncertainty as a value-interval driver and tying the 6-point half-width to small-sample calibration and conservative survey outcome volatility. Accepted the clarity suggestion by stating that the official NBB calendar says 2026-08-21. Kept the specs endpoint check because it was actually fetched during the draft run.

disposition not applicable: Review disposition: accepted the interval critique by removing calendar/source-binding uncertainty as a value-interval driver and tying the 6-point half-width to small-sample calibration and conservative survey outcome volatility. Accepted the clarity suggestion by stating that the official NBB calendar says 2026-08-21. Kept the specs endpoint check because it was actually fetched during the draft run.

disposition not applicable: Review disposition: accepted the interval critique by removing calendar/source-binding uncertainty as a value-interval driver and tying the 6-point half-width to small-sample calibration and conservative survey outcome volatility. Accepted the clarity suggestion by stating that the official NBB calendar says 2026-08-21. Kept the specs endpoint check because it was actually fetched during the draft run.

Activity artifacts

Complete original trace

§Belgium August 2026 NBB Consumer Confidence Forecast

Framing and exact resolver: the target is the National Bank of Belgium's first published August 2026 consumer confidence indicator, seasonally adjusted, in signed index points. The canonical ledger target uses country BE, unit index_points, dataPointId nbb.consumer_confidence.indicator.2026-08.first_print, and an expected release window from 2026-08-16 through 2026-08-24.

▸ Reported tool use: official.lookupmodel report
Fetched the NBB upcoming latest-figures view for consumer confidence releases.
↳ Fetched official schedule numbers: Firms and short-term business statistics, period 2026M8, Consumer confidence indicator, Publication, release date 21 Aug 2026; the same page's August calendar grid shows 2 results for 2026-08-21.
▸ Reported tool use: official.lookupmodel report
Fetched the official NBB July 2026 consumer confidence indicator release page and extracted the headline table.
↳ Fetched NBB release numbers: release date 22 Jul 2026 11:00, period 2026M7; consumer confidence indicator values were 2025 Jul -4, Aug -2, Sep -1, Oct 0, Nov 2, Dec -1, 2026 Jan 4, Feb 1, Mar -6, Apr -9, May -10, Jun -7, Jul -5.
▸ Reported tool use: official.lookupmodel report
Fetched the NBB consumer-confidence methodology page for target identity and construction.
↳ Fetched methodology numbers: the survey covers approximately 1850 households, interviews run between the 1st and 15th days of each month, the indicator is the arithmetic mean of 4 seasonally adjusted key question balances, and balances are bounded from -100 to +100.
▸ Reported tool use: local.modelmodel report
Computed thesis_model_candidate_v1-style one-step benchmarks from the fetched NBB history.
↳ Model candidate numbers from fetched history: persistence point -5.0, p10 -9.3, p50 -5.0, p90 -0.7, 80% interval [-9.3, -0.7], 90% interval [-10.5, 0.5], train cutoff 2026-07, calibration_n 12, walk-forward MAE 2.75; recent monthly changes were 2, 1, 1, 2, -3, 5, -3, -7, -3, -1, 3, 2.
▸ Reported tool use: network.checkmodel report
Checked the requested public specs slug endpoint before finalizing.
↳ Fetched https://app.thesisinstitute.org/specs.json and received HTTP 404 with an HTML body of 11289 bytes rather than a specs JSON document; the registered slug belgium-consumer-confidence-august-2026 was kept.

Base rate/reference class: using the 13 official NBB monthly first prints from 2025-07 through 2026-07, the latest print is -5.0, the 13-print mean is -2.9, the last-six mean is -6.0, the last-three mean is -7.3, and the range is -10.0 to 4.0. Persistence is the base rate because it has the lowest walk-forward MAE among the simple candidates.

Level, momentum, one-off, and policy mechanisms: the level remains negative but has improved from May -10 and June -7 to July -5; momentum is upward, but the July table also shows unemployment expectations at 14, worse than June 13, while savings improved from 19 to 23. I have no direct public August survey signal before the first print, so I do not move materially away from the persistence benchmark.

Prior/update/interval: persistence prior = -5.0 from the July 2026 first print; historical sample = NBB 2025-07 through 2026-07 first-print table; adjustment components = +0.0 for August because July's improvement and unemployment deterioration are already in the latest print and no direct August signal is fetched. For this level survey-balance series, use successive monthly changes: sigma = 3.37 from changes [2, 1, 1, 2, -3, 5, -3, -7, -3, -1, 3, 2]. The raw 80% half-width is 1.28*sigma = 4.31; widen by about 1.39x to 6.0 for small-sample calibration and conservative one-month survey outcome volatility, giving -5.0 - 6.0 = -11.0 and -5.0 + 6.0 = 1.0.

Counter-consideration and falsification: upside risk is a stronger August improvement in household saving and macro expectations, which would land above the interval if the first print is 1.1 or higher. Downside risk is a renewed deterioration in unemployment or financial-situation expectations, which would land below the interval if the first print is -11.1 or lower. Outside the interval would require a larger one-month move than most recent NBB changes.

Review disposition: accepted the interval critique by removing calendar/source-binding uncertainty as a value-interval driver and tying the 6-point half-width to small-sample calibration and conservative survey outcome volatility. Accepted the clarity suggestion by stating that the official NBB calendar says 2026-08-21. Kept the specs endpoint check because it was actually fetched during the draft run.

calibrated forecast · 80% CI
-5.0[-11.0 · +1.0]
Target metadata

Data point: nbb.consumer_confidence.indicator.2026-08.first_print

More government data forecasts