What New Zealand’s adult skills results reveal about acting on capability signals before certainty arrives
Signal Intelligence Briefing 008
Uncertainty should calibrate the response. It should not erase the signal.
New Zealand has received one of its clearest adverse adult-skills signals in years.
The OECD’s 2023 Survey of Adult Skills measured a substantial deterioration in literacy and numeracy compared with New Zealand’s previous participation. Average literacy was 260 points and numeracy 256. Twenty-six per cent of adults were at Level 1 or below in literacy and 28 per cent in numeracy. The OECD describes the declines between 2014–15 and 2022–23 as particularly large in New Zealand and strongest among lower-performing adults.
The result is serious.
It is also imperfect.
New Zealand’s weighted assessment response rate was 48.0 per cent. The OECD’s final technical adjudication recorded Caution–Low for the extended non-response bias analysis and Caution for monitoring and validation. Separate quality checks also identified 301 New Zealand cases associated with interviewers whose response patterns raised concern; the OECD altered how those cases contributed to part of the proficiency-estimation process while retaining their cognitive responses in the estimation of plausible values.
The Ministry of Education therefore had legitimate reasons to resist a literal reading of the headline score changes. Its public account says that “a lot” of the measured change is likely to reflect the difficulty of surveying after COVID-19 and lower respondent engagement, and that equally large actual declines across population groups are “not plausible”.
A separate working paper, Reading Between the Lines, tested those competing interpretations in detail. Its conclusion was deliberately bounded: the survey problems are real and justify caution, but the publicly visible evidence does not quantify how much of the decline they explain. The Ministry’s caution is substantively legitimate; its strongest causal wording goes further than the visible quantification.
That leaves a more consequential question.
What should a capable system do when an adverse signal is strong enough to matter but not precise enough to read literally?
1. Purpose of This Briefing
Primary Goal
To examine how New Zealand should interpret and respond to the deterioration measured by the Survey of Adult Skills when the direction of the signal deserves attention but its precise magnitude and causes remain uncertain.
Core Strategic Premise
Uncertainty changes what action is justified. It does not automatically justify no action.
Good governance does not require pretending that an uncertain measurement is exact. Nor does it require waiting until every uncertainty has disappeared.
The task is to distinguish what the evidence establishes, what remains unknown, what decisions are consequential now, and which responses are proportionate to the confidence available.
What This Briefing Is Not
This is not an argument that New Zealand literally lost the full measured amount of literacy and numeracy capability.
It is not a claim that PIAAC is methodologically clean.
It is not an allegation that the Ministry of Education concealed the results, acted dishonestly or deliberately misled the public.
It is not a claim that PIAAC identifies why adult skills changed.
It is not an argument that government should respond by simply funding more literacy courses.
And it is not a complete redesign proposal for foundation education.
The evidence supports a narrower position: a large deterioration was measured; the exact relationship between that result and underlying national capability remains uncertain; and the uncertainty itself now needs to be governed.
Desired Reader Outcome
Readers should leave able to ask three more useful questions of any consequential capability signal:
What does the evidence justify us believing now?
What action is proportionate to that degree of confidence?
What additional evidence would change the decision?
2. Executive Signal Summary
Signal 1 — A substantial deterioration was measured
The direction of the PIAAC result is clear. New Zealand’s average literacy and numeracy results fell between 2014–15 and 2022–23. The OECD describes the declines as particularly large and strongest among lower-performing adults. The shares of adults at Level 1 or below increased in both literacy and numeracy.
That does not establish the exact amount of underlying capability loss.
It does establish an adverse signal worthy of explanation.
Signal 2 — The measurement concerns are substantive
The case for caution is not rhetorical. New Zealand had a 48.0 per cent weighted assessment response rate. The OECD recorded Caution–Low for extended non-response analysis and Caution for monitoring and validation. The 301 interviewer-linked cases required special technical treatment.
A literal reading of the score changes would therefore be too confident.
Signal 3 — The alternative explanation remains incompletely quantified
The Ministry argues that survey conditions and lower respondent engagement probably explain a substantial share of the measured change.
That is plausible.
But plausibility and magnitude are different claims. The public technical record establishes credible reasons why the survey may overstate the underlying decline. It does not provide a visible decomposition showing how much of the decline those mechanisms explain.
The appropriate conclusion is therefore uncertainty about magnitude, not certainty that the adverse signal disappears.
Signal 4 — New Zealand already has intervention infrastructure
The system is not starting from zero.
In 2026, the Intensive Literacy and Numeracy Fund remains active. TEO-led and employer-led workplace literacy and numeracy funding also remain available. Eligibility has widened in some areas, including additional access for New Zealand apprentices and some learners studying at Level 4 or above where literacy or numeracy needs are demonstrated.
Those changes should not be attributed to PIAAC without evidence of causation. Their significance here is simpler: New Zealand already has mechanisms through which targeted, bounded action can occur.
Signal 5 — The deeper gap may be decision architecture
The PIAAC dispute reveals a structural problem larger than one survey.
National measurement can produce estimates, caveats and competing interpretations. The harder question comes next:
When is the evidence strong enough to justify action, what kind of action should uncertainty permit, and how should that action generate better evidence?
A mature capability system needs a loop between measurement, judgement, intervention and learning.
3. Why This Matters
Capability risk does not disappear because the estimate is uncertain
Around a quarter of New Zealand adults scored at Level 1 or below in literacy and more than a quarter did so in numeracy. PIAAC measures adults’ ability to engage with written information, mathematical information and problem-solving tasks across contemporary contexts. The OECD treats these capabilities as foundations for learning, work and participation in information-rich environments.
A system can reasonably debate how much conditions changed between survey cycles while still asking whether the present distribution of capability warrants attention.
Those are different questions.
The lower end matters more than the mean alone
National averages compress distributions.
That matters here because the OECD found that the deterioration was strongest among lower-performing adults.
Even if the national mean ultimately proves to have fallen by less than the raw comparison suggests, concentration of weakness at the lower end may still carry significant consequences for learners, workers, employers and communities.
The practical question is therefore not only how far the mean moved. It is where capability risk is concentrated and how well current systems can see and respond to it.
Waiting is also a decision
There is a sound instinct towards caution when evidence is disputed.
The problem begins when “do not overinterpret this evidence” silently becomes “there is not yet enough evidence to do anything”.
Those propositions are not equivalent.
The costs of acting too strongly are real. Government can direct resources towards the wrong problem, redesign funding around a misleading indicator or launch a programme whose theory of change is weak.
But delay also has consequences. Existing funding rules, provider incentives, assessment practices and access pathways continue operating whether the national interpretation has been resolved or not.
The governance problem is therefore not action versus caution.
It is how to design action that remains proportionate to uncertainty.
4. Current System Reality
The deterioration is real as a measured result
The Survey of Adult Skills directly assessed New Zealand adults aged 16 to 65 in literacy, numeracy and adaptive problem solving.
In 2022–23, average literacy was 260 points and numeracy 256. Compared with 2014–15, both declined. The OECD describes the changes as particularly large in New Zealand and reports that lower-performing adults deteriorated most.
Those are observations from the survey.
They are not yet a causal explanation.
The survey operation creates legitimate uncertainty
The OECD’s final technical report records a weighted background-questionnaire response rate of 49.3 per cent and a weighted assessment response rate of 48.0 per cent for New Zealand. The extended non-response bias analysis was rated Caution–Low, meaning potential non-response bias remains a material interpretive consideration even though several checks did not show notable changes in mean proficiency estimates.
Monitoring and validation also received a Caution rating. The OECD records high interviewer attrition and incomplete achievement of the validation threshold.
Separately, the OECD identified unusual cognitive-response patterns clustered around a small number of interviewers. For New Zealand, 301 cases from identified interviewers were excluded from the data used to estimate the population model, while their cognitive assessment responses still contributed to the estimation of plausible values. The OECD states that there was no definitive evidence of data falsification or interviewer misconduct across those cases.
These findings prevent a clean equation:
measured score decline = identical decline in underlying national capability.
But measurement uncertainty does not provide an alternative number
This is the hinge.
The Ministry’s interpretation is that post-COVID survey difficulty and lower respondent engagement probably account for a substantial part of the deterioration. Its public summary says that “a lot” of the change is likely due to those conditions and that equally large actual declines across population groups are not plausible.
That explanation may be right.
The visible evidence examined in the working paper does not establish its magnitude.
There is no contradiction in saying both:
the raw score change should not be treated literally;
and
the direction and distribution of the result should still be taken seriously.
The system already contains pieces of a response
TEC’s Intensive Literacy and Numeracy Fund remains active in 2026. Its stated purposes include raising adult literacy and numeracy, increasing opportunities to participate, and improving the quality and relevance of provision, including the ability to identify learner need and learning gain.
TEO-led and employer-led Workplace Literacy and Numeracy funds provide additional routes in paid-work contexts. In 2026, both include widened eligibility for some New Zealand apprentices and learners studying at Level 4 or above, subject to demonstrated literacy or numeracy need and other funding conditions.
Separately, released Ministry policy work states that the fundamental purpose and parameters of foundation education policy settings have not been systematically reviewed for many years or updated to reflect new evidence. It argues that short-term improvements alone will not address underlying problems associated with declining participation and adult literacy and numeracy, and identifies broader questions around policy settings, funding, pathways and connections to employment.
These developments should not be collapsed into one causal story.
The defensible observation is that response infrastructure exists, while parts of the policy system are already reconsidering the wider architecture.
5. What the System May Be Misunderstanding
Misunderstanding 1 — Uncertain magnitude means uncertain relevance
It does not.
Uncertainty about magnitude should change the confidence and scale of a response. It does not automatically remove the decision problem.
A national capability signal can be too imprecise to support a precise claim and still be strong enough to justify further diagnosis, targeted testing or bounded intervention.
Misunderstanding 2 — Methodological caution makes the status quo neutral
The status quo is not neutral.
It contains funding rules, eligibility thresholds, provider incentives, assessment practices and participation patterns. Choosing not to change them is still a decision about where attention and resources remain.
The relevant comparison is therefore not intervention versus doing nothing.
It is one set of consequences versus another under conditions of uncertainty.
Misunderstanding 3 — A national survey is a local diagnosis
PIAAC can identify national and subgroup patterns. It cannot tell a provider what a particular learner needs, or an employer what a particular workforce problem is.
Operational decisions require evidence closer to the person and context.
National measurement should therefore guide where to look more closely, not substitute for local diagnosis.
Misunderstanding 4 — Acting means creating another national programme
It does not.
New Zealand already has targeted intervention routes. Before adding new machinery, the system can ask whether existing mechanisms are reaching the right people, whether their eligibility assumptions remain useful, and whether evidence of learning gain is strong enough to support future decisions.
Misunderstanding 5 — Measurement is mainly reporting infrastructure
This may be the deeper issue.
National measurement often culminates in a report, a briefing and a public argument over the headline.
But the more useful question is what happens between measurement cycles.
Capability measurement should function as decision infrastructure.
Its purpose should be to improve decisions about what to test, where to intervene, what to stop, what to scale and which evidence should be gathered next.
6. Emerging Adaptation Patterns
There is not yet evidence of one coherent national response to the PIAAC deterioration.
There are, however, several relevant movements.
Pattern A — Foundation education is being reconsidered as a system problem
Released Ministry advice argues that foundation education policy settings have not been systematically reviewed for many years and proposes broader consideration of policy settings, funding, pathways through education and connections to employment and other interventions.
This is policy work, not an implemented redesign.
Its significance is that the problem is being framed beyond a single course or learner deficit.
Pattern B — Sector actors are calling for greater coordination
Ako Aotearoa’s Manako programme published a series of 2025 briefings on PIAAC and adult literacy and numeracy. One proposes a new system-wide Language, Literacy and Numeracy Implementation Strategy for tertiary education. The briefing papers were sent to the Minister for Tertiary Education and Skills and other stakeholders.
This is a sector proposition rather than government policy.
But it is evidence that the PIAAC results have reopened questions about system coherence.
Pattern C — Existing targeted funding remains operational
The Intensive Literacy and Numeracy Fund and both workplace literacy and numeracy routes remain active in 2026. TEC requires evidence of low literacy or numeracy need for workplace eligibility, including use of the Literacy and Numeracy for Adults Assessment Tool in the relevant pathways.
This creates an existing base from which the system can learn.
National survey evidence does not have to carry the full diagnostic burden if operational evidence is used deliberately alongside it.
Pattern D — The policy problem is moving towards connection
Across the Ministry and sector material, a recurring question is how foundation capability connects to further study, employment, funding settings and the rest of the education system.
That is still an emerging direction rather than a settled architecture.
But it suggests that the next phase of adult literacy and numeracy may be less about finding one programme that fixes the problem and more about improving the connective infrastructure around capability development.
7. Strategic Implications
For government and system leaders
The immediate requirement is not another declaration of crisis.
It is a decision framework.
For material capability signals, government should be able to state what has been observed, what confidence exists in the measurement, what causal explanations remain plausible, which decisions cannot reasonably wait, which interventions are reversible, what evidence those interventions should produce, and what future findings would trigger expansion, redesign or withdrawal.
That would allow uncertainty to become part of governance rather than an excuse for either overreaction or drift.
For TEC and funders
The PIAAC results should prompt questions about the intelligence generated by current investment.
Not only: How many learners participated?
But: Who was reached? What capability changed? How durable was that change? Where is provision failing to reach demonstrated need? Which settings produce better outcomes for which people?
The current funding architecture already refers to learner need and learning gain.
The strategic opportunity is to make that information more useful as system intelligence.
National survey evidence and operational learner evidence should triangulate rather than compete.
For providers and educators
A national score does not diagnose an individual learner.
The practical task remains identifying actual capability, understanding the context in which it needs to be used, designing relevant learning and gathering credible evidence of progress.
The PIAAC signal raises the evidential standard as much as it raises the policy question.
If public investment is intended to improve adult capability, the system needs more than evidence that teaching occurred.
It needs evidence that capability developed.
For employers and workforce systems
Literacy and numeracy are not only education-sector concerns.
TEC’s workplace funding guidance already recognises that low literacy or numeracy can appear through difficulties with instructions, forms, reporting, calculations and digital devices required for work.
That makes workplaces both part of the problem context and part of the evidence environment.
Employers may hold useful information about where foundational capability creates friction, what support changes performance and whether gains transfer into real tasks.
8. Intervention Options
There is no single response demanded by the evidence.
Option 1 — Wait for stronger national evidence
Hold substantial policy change until further analysis or another major data source provides stronger confidence about the direction and magnitude of change.
Suitable when proposed interventions are expensive, highly disruptive or difficult to reverse.
Strength: reduces the risk of redesigning the system around a misleading estimate.
Trade-off: delay becomes the default response even if a meaningful underlying deterioration is genuine.
Option 2 — Prioritise measurement and triangulation
Invest first in resolving the evidential questions through further PIAAC analysis, operational assessment data, provider and workplace evidence, and future national or cohort measures.
Suitable when the principal barrier to action is genuinely uncertainty about who is affected, where the problem sits or what should be targeted.
Strength: improves diagnosis and future targeting.
Trade-off: better measurement does not itself build capability.
Option 3 — Take targeted low-regret action through existing infrastructure
Use existing ILN and workplace literacy and numeracy mechanisms to test targeted responses while strengthening evidence about need, participation and learning gain.
Suitable when evidence of need is stronger locally than national causal certainty and interventions are reasonably reversible.
Strength: action can generate additional evidence rather than waiting for evidence to arrive from elsewhere.
Trade-off: local improvement can remain fragmented if the wider system architecture is the underlying constraint.
Option 4 — Redesign the wider foundation-capability system
Treat PIAAC as one of several signals justifying a broader review of foundation capability, including funding, pathways, provider sustainability, workplace connection, educator capability, assessment and evidence of outcomes.
Suitable when evidence suggests the problem lies not primarily in individual programmes but in how the system fits together.
Strength: addresses fragmentation and structural constraints.
Trade-off: a large reform process can consume substantial time and attention before producing better outcomes and may attribute too much diagnostic power to PIAAC.
Option 5 — Improve the intelligence system while taking bounded action
Combine stronger measurement with targeted intervention.
Use national evidence to identify risk. Use local and operational evidence to select interventions. Use intervention data to refine the system’s understanding. Escalate only when accumulating evidence justifies greater commitment.
This is the provisional direction that appears most consistent with the current evidence.
It preserves caution without making caution passive.
9. Emerging Directional Principles
1. Calibrate action to uncertainty; do not equate uncertainty with inaction.
2. Separate the measured change, measurement uncertainty and causal explanation.
3. Demand stronger evidence for stronger causal claims.
4. Raise the evidential threshold as decisions become more expensive, consequential or irreversible.
5. Prefer reversible, evidence-generating interventions while important uncertainty remains.
6. Pay attention to the distribution of capability, not only the national mean.
7. Triangulate national measures with evidence closer to learners, workplaces and providers.
8. Use existing capability infrastructure before creating new machinery.
9. Treat activity, participation and capability improvement as different things.
10. Treat measurement as decision infrastructure rather than periodic reporting.
10. Watchlist
The interpretation in this briefing should remain revisable.
Watch for:
- a credible New Zealand-specific sensitivity analysis that quantifies how much non-response, interviewer-linked cases or engagement affected the national estimates;
- a counterfactual analysis showing materially different literacy or numeracy scores under alternative treatment of the interviewer-linked cases;
- release-era Ministry or OECD analysis that supplies a quantitative basis for the claim that a large share of the measured decline was survey-generated;
- future OECD or Ministry clarification that materially changes the interpretation established in the working paper;
- PISA and other school-level evidence as partial cohort triangulation, while recognising that school assessment is not a substitute for adult-skills measurement;
- trends from the Literacy and Numeracy for Adults Assessment Tool or other operational measures that corroborate, complicate or contradict the national direction;
- evidence about who currently accesses ILN and workplace provision, who does not, and where learning gain is strongest;
- government decisions arising from foundation-education policy work;
- evidence that the lower-proficiency tail is stabilising or improving; and
- any credible counter-signal showing that the apparent deterioration is predominantly an artefact of measurement rather than meaningful capability change.
If stronger evidence eventually shows that the survey substantially overstated the underlying decline, the interpretation should become less alarmed.
That is not a failure of the argument.
It is what an evidence-responsive system is supposed to do.
11. Conclusion
New Zealand does not know exactly how far adult literacy and numeracy capability has fallen.
That matters.
So does the fact that PIAAC measured a large deterioration, particularly among lower-performing adults.
The survey operation had real weaknesses. The Ministry was right to resist a literal reading of the headline numbers. Its stronger claim that survey conditions explain a large share of the decline remains plausible but not publicly quantified.
Those propositions can coexist.
The mistake would be forcing them into a binary:
either the measured decline is literally true, or there is no meaningful signal;
either the Ministry’s explanation is wrong, or the survey can be dismissed;
either government launches a national response, or waits for better data.
A mature capability system should be able to hold a more difficult position.
We do not yet know exactly how large this problem is.
We know enough to investigate it seriously.
We already possess mechanisms through which bounded action is possible.
Those actions can generate better evidence.
And the system should be willing to change course when the evidence changes.
That is not a compromise between action and caution.
It is what caution looks like when it has to govern.
The deeper shift is from treating national measurement as a periodic description of capability to treating it as part of an ongoing capability-intelligence system.
Measurement should not end with the report.
It should improve the next decision.
12. References
OECD. Survey of Adults Skills 2023: New Zealand. 10 December 2024.
OECD. Survey of Adult Skills 2023 Technical Report. OECD Skills Studies. 8 August 2025.
OECD. Annex G — Data adjudication in PIAAC. Survey of Adult Skills 2023 Technical Report. 2025.
Ako Aotearoa / Manako. Language, literacy and numeracy briefing papers. 2025.
Tertiary Education Commission. Intensive Literacy and Numeracy Fund. Updated 28 January 2026.
13. Key Questions
Did New Zealand’s adult literacy and numeracy really decline?
PIAAC measured a substantial decline in both literacy and numeracy compared with New Zealand’s previous participation. That measured direction is clear. What remains uncertain is how closely the size of the measured change corresponds to an equivalent decline in underlying national capability, because response, engagement and fieldwork concerns affect interpretation.
How much confidence should we place in the PIAAC numbers?
Enough to take them seriously, but not literally.
The survey provides meaningful evidence about direction, distribution and current proficiency. New Zealand’s data also carry material quality qualifications. The appropriate response is to preserve both facts rather than choosing one.
Could survey problems explain much of the decline?
Possibly.
The Ministry believes they do, and the evidence provides credible mechanisms through which the measured decline could have been overstated. What remains unshown publicly is the quantitative bridge between those mechanisms and the claim that they explain a large share of the decline.
Should New Zealand act before the exact magnitude is known?
That depends on the action.
Expensive, irreversible system changes require stronger evidence than bounded pilots, improved targeting, better measurement or the deliberate use of existing intervention infrastructure.
The principle is proportionality: uncertainty should affect the scale and reversibility of the response rather than automatically preventing one.
What would a proportionate response look like?
Strengthen the intelligence system while taking bounded action.
Use national evidence to identify material risk. Use local, learner and workplace evidence to target interventions. Prefer existing mechanisms where they are fit for purpose. Measure what changes. Escalate, redesign or stop as evidence accumulates.
Does this mean New Zealand needs a new national literacy and numeracy strategy?
Not necessarily.
A coordinated strategy is one credible option and has already been proposed by sector actors. The evidence in this briefing does not establish that one particular national architecture is the answer.
The first question is what problem a new strategy would solve that current arrangements cannot.
14. Continue Exploring
Reading Between the Lines — What New Zealand’s PIAAC Results Show — and What the Official Explanation Does Not Yet Establish
The full evidence analysis underpinning this briefing.
SIGNAL 007 — When Output Is No Longer Evidence
SIGNAL 007 examined the inferential problem at the level of individual capability: when a visible output can no longer carry the full claim, institutions need better evidence architecture.
SIGNAL 008 extends the same discipline to system evidence.
SIGNAL 006 — The Human Operating Envelope
15. Subscribe
Signal Intelligence examines structural changes in AI, capability, education, assessment, trust, workforce systems and institutional design.
16. Tags
capability; literacy and numeracy; PIAAC; adult skills; evidence; uncertainty; governance; workforce; education; foundation education; capability intelligence; human judgement; Signal Intelligence