EXPOSED
The mechanical core of the job — writing R/SAS/Python analysis scripts, running regressions and survival models, cleaning datasets, generating tables and figures, drafting methods sections — is exactly what current models do at usable quality, and fast. What resists is upstream and downstream: designing an experiment or sampling frame that can actually answer the question, defending assumptions when the data violate them, and putting your name on an inference that a regulator, court, or executive will act on. The modal statistician (biostatistics, government survey work, market research support) still spends a large share of hours on the automatable middle, which is why this lands squarely in EXPOSED rather than safe.
This fall is concentrated in 2020 and has not recovered since.
Median pay $91,160 → $105,650 -7.3% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
+8.5% 32,200 → 34,900 on the projections basis
Exposed, but growing
AI can already do a lot of these tasks, and the BLS still expects +8.5% more of these jobs by 2034. Demand for the output is growing faster than the work is being automated away — the mechanism BLS gives for software developers, and the combination people most often misread as an error.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~2,000 openings a year on average, including replacing people who leave.
DemographerBiometricianData AnalystData ManagerData ModelerStatisticianData EngineerBiostatisticianData SpecialistSampling ExpertBiomathematicianClinical AnalystData CoordinatorDatabase AnalystApplied ScientistFinancial AnalystResearch ScientistTrend InvestigatorPostdoctoral FellowSports StatisticianStatistical AnalystSurvey StatisticianApplied StatisticianStatistical Engineer
Holding it up: judgment & accountability . Weakest point: embodiment .
Mixed — a routine tier and a judgment tier A 9 reflects the split: model fitting, multiple-imputation runs, table shells, and CDISC/ADaM dataset production are already machine-generatable, while sample-size justification for a Phase III protocol, choosing a variance estimator for a complex multistage survey design, and diagnosing why a proportional-hazards assumption fails in the data you actually have still require someone who understands the study, not just the file.
Fully desk- and screen-based A 2 rather than 0 only because some statisticians walk to a lab or clinic to see how measurements are actually recorded, or sit in on a data-monitoring committee in person; the work itself is a laptop, a compute cluster, and a version-controlled repo.
Certification preferred, not legally required At 5 you have credentials that matter socially — a PhD, ASA Accredited Statistician, SAS/ASQ certifications — but no state licence bars anyone else from running the analysis, and when an FDA submission or a court's Daubert challenge goes badly the sponsor, the PI, or the testifying expert absorbs it rather than the staff statistician who wrote the SAP.
Meaningful discretion A 13 is earned by calls that have no procedure — declaring an interim futility boundary crossed, deciding whether nonresponse in a survey wave is ignorable, choosing whether to pool sites with sparse events — decisions a reviewer will second-guess and that move drug approvals or published federal estimates, but stopping short of 14+ because most of those calls are pre-specified in an SAP or a documented methodology, reviewed by a senior statistician, and signed off by someone above you.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (9/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (5/20) is whether the law requires a licensed human to sign. Trust premium (8/20) is whether buyers specifically pay for a person. Judgment and accountability (13/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 26 of this occupation's 37 points (70%).
Embodiment (2/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 58/100, still EXPOSED.
Genuine two-tier job: if scripting, table generation, and methods drafting are fully automated, the remaining hours concentrate on design of sampling frames, estimand definition, adjudicating assumption violations, and pre-registration defense — the tier models are worst at. Raises measured resistance without any new law, but only for statisticians already doing design work; shrinks headcount at the same time
Court adoption of Daubert-style rules that statistical expert testimony must come from a human affiant who can be deposed on method choices, explicitly barring AI-generated analyses as the sole basis — Federal Rule of Evidence 702 amendments (2023) already tightened proponent burden on reliability
FDA/ICH-style requirement that a named, credentialed trial statistician personally attests to the statistical analysis plan and unblinded analysis — already partly real via ICH E9(R1) estimand attestation and 21 CFR Part 11 e-signature on SAPs; extension to explicit personal sign-off on any AI-generated analysis, plus EMA/FDA guidance requiring human attestation of model-assisted submissions, would harden this
Formalization of the statistician as the accountable party for reproducibility and p-hacking failures — e.g., journals or NIH requiring a named statistical reviewer of record on submissions, as NEJM and some FDA advisory panels already do informally
Litigation and regulatory buyers specifically contracting for human-authored analysis because AI-generated evidence faces admissibility or audit challenge — visible already in expert-witness engagement letters requiring disclosure of AI use
ASA/RSS establishing an accredited chartered-statistician credential that regulators or IRBs require for sign-off on human-subjects analyses (RSS CStat exists but carries no statutory force in the US)
The limit. Even with all levers, the sign-off tier is a small fraction of the 29k headcount; liability shields protect roles, not the volume of routine analyst-hours, so the occupation likely shrinks toward a smaller, better-protected core rather than holding its size.
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 3,220 | $136,660 +29% |
| Seattle-Tacoma-Bellevue, WA | 2,560 | $105,650 +0% |
| Boston-Cambridge-Newton, MA-NH | 1,950 | $102,320 -3% |
| New York-Newark-Jersey City, NY-NJ | 1,460 | $137,510 +30% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 920 | $105,820 +0% |
| San Francisco-Oakland-Fremont, CA | 800 | — |
| Durham-Chapel Hill, NC | 600 | $103,290 -2% |
| Los Angeles-Long Beach-Anaheim, CA | 580 | $116,200 +10% |
| San Jose-Sunnyvale-Santa Clara, CA | 320 | $202,840 +92% |
| Raleigh-Cary, NC | 210 | $152,880 +45% |
| Boulder, CO | 100 | $144,990 +37% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 37. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.