# The Politics of Official Statistics Official statistics are infrastructure. Markets price off them, benefits index to them, appropriations depend on them, and the public evaluates governments by them. That makes the agencies producing them a natural target — not because the arithmetic is easy to falsify, but because **the credibility of the number is a political asset, and destroying credibility is far easier than manufacturing false data.** ## Three distinct attack surfaces **Contest the definition.** Change what counts. Which unemployment measure is headlined; whether discouraged workers are included; which price index governs benefits (see [[Index Choice and Chained CPI]]); what the [[GDP as a Constructed Measure]] boundary includes. These fights are legitimate on their face and consequential in effect. **Contest the collection.** Underfund the survey, cut the sample, reduce collection frequency, add questions that suppress response. A degraded statistic is noisier, and noise is deniability — every unwelcome reading can be attributed to measurement error. Response rates to major federal surveys have declined for years, and reduced sample sizes raise variance in exactly the sub-national and demographic detail that policy depends on. **Contest the producer.** Replace or discipline the personnel. This is the most direct route and the one with the shortest lag. ## The August 2025 rupture The [[Bureau of Labor Statistics]] had been among the most insulated statistical agencies in the world since 1884. On **August 1, 2025**, Trump fired Commissioner [[Erika McEntarfer]] hours after a weak jobs report — the precipitating event of what this vault tracks as the [[BLS Data Integrity Crisis]], which then ran through the [[E.J. Antoni]] nomination and withdrawal, a shutdown-driven data blackout, and the [[Brett Matsumoto]] nomination. The mechanism is worth stating precisely, because the damage does not require any number to be falsified. **Once the head of a statistical agency can be dismissed for publishing an unwelcome figure, every subsequent figure is read through that fact.** Nothing has to change in the methodology. Market participants, researchers, and the public now have to price in the possibility of interference, and there is no observation that can rule it out. Statistical credibility is a stock built over more than a century of not doing this, and it is depleted by a single instance. The second-order effect is worse than the first. A rate-sensitive economy runs on the assumption that [[Nonfarm Payrolls]], the [[Unemployment Rate]], and the [[Consumer Price Index]] are produced by people insulated from the political consequences of what they find. Remove the insulation and the numbers keep coming, but their function as a shared factual baseline degrades. ## Why "just falsify the data" is the wrong model Direct fabrication is rare, detectable, and unnecessary. Federal statistics are produced by large professional staffs, published on pre-announced schedules with documented methodology, and cross-checked against private-sector series. Crude manipulation would surface quickly. The productive attacks are the ones that never require a false number: - **Definitional revision**, which is legal and defensible. - **Framing selection**, choosing which true statistic to headline — see [[Framing]] and [[Rate Versus Level]]. - **Credibility attack**, which does not touch the data at all and works by making the audience discount it. This is [[Manufacturing Doubt]] applied to measurement, and it is the cheapest of the three. ## The asymmetric payoff Undermining official statistics is unusually attractive because the benefit is immediate and the cost is deferred and diffuse. A government that discredits the unemployment series gains relief from an unwelcome news cycle now. The loss — an economy without a trusted common measurement layer — arrives later and is borne collectively. It also does something more specific: **it removes the possibility of being proved wrong by evidence.** A polity without credible shared statistics cannot settle empirical disputes, which converts every factual question into a question of which side you are on. That is not a side effect of attacking statistical agencies; for some actors it is the objective. ## Related For the operational chronology of the 2025–2026 statistical-agency fight, see [[BLS Data Integrity Crisis]] — this page covers the mechanism, that page covers the events. [[Metric Capture]] · [[GDP as a Constructed Measure]] · [[Index Choice and Chained CPI]] · [[Rate Versus Level]] · Goodhart's Law · [[Manufacturing Doubt]] · [[Bureau of Labor Statistics]] · [[Mathematics and Power Hub]]