Teaching children history and civics

The program (what a school should actually do)

History and civics carry more curricular argument per unit of evidence than any subject in this archive. England rewrote a national curriculum around knowledge; American states legislated civics tests; a generation of teacher education was reorganised around "historical thinking." Not one of those three decisions has a causal evaluation behind it.

What does exist is a small, surprisingly good set of randomised trials, and they agree with each other in an inconvenient way: everything in this subject works on the thing it teaches and nothing works on anything else.

Ages 5–14 — build the knowledge, cumulatively, and budget in years.

• Run history as a coherent, text-heavy, content-rich curriculum, sequenced across years rather than assembled from units. This is the same bet as curriculum choice's highest-upside recommendation and background knowledge's durable lever. (evidence: knowledge vs skills)
• Expect history knowledge, not reading comprehension. Across eight randomised trials, content-rich history instruction moved history knowledge at g = 0.19–0.46 and moved standardized reading comprehension in exactly one trial — which the same team failed to replicate twice.
• If reading comprehension is the goal, the dose is years. A three-year content spiral gives 0.11–0.12 on standardized reading and holds at 14 months; a six-year whole-school Core Knowledge lottery gives ITT 0.24 on state tests. Nothing shorter has ever moved a standardized measure.
• Do not expect it to close gaps. Content-rich instruction compounds prior knowledge: the kindergarten effect ran 0.53 at −1 SD entering vocabulary and 1.31 at +1 SD, and low-prior-attainment students in one high-school trial got no benefit at all.

Ages 13–18 — teach with documents, for the documents.

• Give students multiple primary sources instead of a textbook. In the one factorial that isolated the ingredients, the documents raised content knowledge and sourcing behaviour; the explicit heuristic instruction bolted on top did not raise content. (evidence: historical source work)
• Adapt the sources. About a third of teacher modifications to the leading document-based curriculum are made for reading access. A student who cannot read the source cannot source it.
• Protect whole-class text discussion, and pay for real training. The one evaluation that produced effects gave teachers four days of summer PD plus two follow-up workshops — and still only three of five attempted discussion, which the evaluator names as the likely reason the higher-order skills never moved.
• Do not buy source work as a literacy intervention. Seven trials of the leading programme give g = 0.14 ns on standardized comprehension.

Ages 14–18 — teach civics for knowledge, and teach the procedure.

• Teach civic knowledge, repeatedly. A 21-lesson randomised curriculum moved civil-liberties knowledge d = 0.49 — and that gain had decayed to the control mean within two years, with zero transfer to general political knowledge. Cumulative repetition is the only plausible response. (evidence: civic education)
• Teach the procedure of participating: registration, deadlines, where and how to vote. Actionable procedural instruction is the one civics-adjacent thing with a positive signal, and it costs an afternoon.
• Do not promise attitudes, tolerance or turnout. All three have been tested under randomisation. All three failed.

What to refuse to spend money or time on

• Generic "historical thinking skills" as the organising principle of the curriculum. The one study that randomly selected which teachers got the training found nothing on its own aligned writing measure; the apparent effect lived entirely among teachers who volunteered for a second year.
• Content-rich history sold as a one-year reading intervention. Eight trials, one positive, failed replication. And the cleanest demonstration in the archive of what an aligned measure buys: the same kindergarteners, same semester, aligned social-studies knowledge +0.93 SD, standardized Woodcock-Johnson Social Studies 0.00 SD, p = .999.
• Deliberation and "open classroom climate" packages. The field's most-cited positive finding collapsed when someone held content, texts and time constant — students genuinely perceived a more open climate, and knowledge, interest, values and discussion all moved zero.
• Mandated community service. Maryland's requirement raised 8th-grade service and later cohorts did less voluntary activity. Sibling-fixed-effects analysis finds service-learning coursework null on adult turnout.
• State civics-test mandates as a turnout policy. At most 1.5 percentage points, not significant, with a marginally significant −9.7 pp among Black youth that nobody has explained.
• Live guest eyewitnesses as the route to critical source reading. In a randomised head-to-head, students who met a live witness scored lower on understanding oral history and on recognising a source as constructed than students given the same interview on video or as a transcript — while being more convinced they had learned more.
• Teacher enthusiasm as evidence. In the 60-school null of a document-analysis history curriculum, significantly more treatment teachers agreed they were better able to build historical perspective-taking. Nothing had improved. This archive has now recorded that pattern three times.
• A civics-focused school mission as a route to voting. The charter lottery that raised voter-file-validated turnout by 6 points found civics-mission schools no better than the rest (p = 0.391) — despite those schools demonstrably raising AP civics coursework.

The measurement problem in this subject

Two of the archive's cleanest measurement demonstrations live here, and a founder who internalises them will read every history and civics claim correctly for the rest of their career.

The alignment demonstration. Same children, same semester, same construct, two instruments: curriculum-aligned social studies knowledge 0.93 SD; the standardized subtest of the same construct 0.00 SD (p = .999). The one apparently significant standardized effect — a vocabulary subtest at 0.09 — collapsed to 0.01 (p = .702) once the authors removed three items whose words (canoe, moccasins, loom) the curriculum explicitly taught. METHODOLOGY's nominal correction for researcher-designed measures is 2×. At a one-semester dose here it is not a discount; it is the entire effect.

The independence demonstration. In the document-based-history cluster, nine of thirteen included sources are developer-led or developer-involved. The two genuinely independent, school-randomised trials are the two nulls. That is the efficacy-to-effectiveness decay curve visible in a single scatter of thirteen studies.

And in civics, the outcome-class demonstration. Knowledge is a domain-skill. Voting and volunteering are behaviour. Tolerance, efficacy and interest are non-cognitive. The literature moves between them freely and the evidence does not: the curriculum that raised knowledge d = 0.49 moved tolerance −0.071, and the school-level treatments that raised validated turnout by 6 to 12 points did it without any civics content advantage at all.

The hereditarian bottom line for a founder

1. The gap-closing promise fails here, and it fails in the direction the premise predicts. Knowledge-rich teaching compounds prior knowledge — the effect ran 0.53 at low entering vocabulary and 1.31 at high, English learners 0.58 against native speakers 0.99, and low-pretest students in one trial gained nothing. This is what METHODOLOGY predicts when instruction removes an environmental bottleneck: the ceiling rises and the spread widens. It is evidence the teaching worked, and it is a reason not to sell it as equity policy.
2. Civics has the archive's cleanest direct test of the premise. Monozygotic twins discordant in education show that genes and shared family substantially confound the raw education–participation gradient — the correlation the entire observational civic-education literature rests on. Within-pair effects survive in the US and Denmark and are null in Sweden.
3. Watch which studies use self-report. The large American education→voting effects are self-reported voting, and educated people over-report voting more. The studies with validated outcomes — voter files, digitised Swedish election rolls — return nulls or credit something other than civics.
4. Every positive result in civics is a compensation effect. Sweden's bottom SES quartile +11.62 pp against a null average; knowledge gains concentrated in Latino and immigrant students; turnout effects largest where baseline participation was lowest. Where civic education does anything, it does it for the children who arrive with least — the mirror image of the history-knowledge pattern, and worth sitting with.

Nothing in this subject bears on g.

Confidence

Open questions a founder should watch

• Is the Core Knowledge lottery replicable? It is the archive's strongest curriculum result, it is still an unpublished working paper three years on, it has no critique and no replication, and it introduces a novel bias correction of the authors' own devising for "two previously undiscovered sources of bias inherent in kindergarten lotteries." Nobody has vetted it.
• Why did the three-year content spiral move mathematics as much as reading? Either the mechanism is not knowledge transfer or knowledge transfer is not specific. Both readings change the recommendation.
• Where between one semester (0.00) and six years (0.24) does knowledge start showing on a standardized test? Nobody has mapped the dose–response curve, and it is the single most decision-relevant unknown in the subject.
• Chronological or thematic? Unevidenced. There is no causal literature on how to order a history curriculum, despite a century of confident argument.
• Can contextualization or corroboration be taught at all? Four independent studies measured them; none moved them.
• Why do good schools produce voters? The mechanism analysis points at noncognitive follow-through rather than knowledge, and nobody has isolated what inside a high-performing school does it.
• No causal evaluation of England's 2014 knowledge-led national curriculum exists — a whole national curriculum rewritten around this subject's central question, leaving no identification strategy behind.
• An independent evaluation of the leading document-based history curriculum is under way (a US Department of Education-funded AIR study). It is the single thing most likely to change a verdict here.