What the NSS Won’t Say About Itself

This morning, the Office for Students published the results of the National Student Survey 2026. This is the first release day since I recently stepped away from that responsibility that I have not spent refreshing a portal that might arrive late, monitoring response rates against the publication threshold subject by subject, and fielding a vice-chancellor hovering at my shoulder asking why the numbers were not yet on his desk (before the results were even published by the OfS…).

Seven iterations of managing NSS data collection from the institutional side leave a residue that does not disappear simply because the job has changed — not that I ever took a publication day to be a neutral reporting exercise in the first place; the process was never one without its influences.

I remember occasions when certain colleagues seemed quietly relieved rather than disappointed that a small subject area had failed to clear the response-rate threshold required for its results to be published separately: an outcome that had nothing to do with the quality of anyone’s teaching, and everything to do with which figures a department would rather not have attached to its name.

By lunchtime, most of what can usefully be said about those results, in the narrow sense of which numbers moved and by how much, will already have been said — WonkHE will have its usual sharp, fast turnaround on the headline movements; the trade press will have its “winners and losers” tables; a handful of vice-chancellors will have issued statements about their institution’s “outstanding” performance before the ink on the release is dry.

None of that coverage is wrong, exactly. But almost all of it will treat the NSS as if it were simply a scoreboard — a set of numbers that rise and fall in response to teaching quality, and can therefore be read off directly as evidence of it. In Satisfied? Student Surveys, Social Position, and the Limits of Experience Data, published here in March, I argued that this reading rests on a category error: satisfaction is not learning, and the gap between the two is not random noise but a socially patterned effect. Publication day is a useful occasion to make a different argument, one that has nothing to do with this year’s specific numbers and everything to do with what the instrument has become: institutions are not passive subjects of this measure. They act on it, in at least three ways worth separating out: in how resource gets allocated toward what the survey happens to ask about, in the choreography of how and when results are announced, and in the survey’s uneasy extension into contested political terrain it has no capacity to adjudicate.

A recap, and a qualification

The March piece leaned on a Bourdieusian account: satisfaction is a judgement made from a position in social space, shaped by a habitus formed well before a student arrives at university (Bourdieu, 1984). On this account, a student whose prior education calibrated them to expect little may report high satisfaction with provision that a student with different expectations would rate as mediocre. Reay, Crozier and Clayton (2009) found something like this dynamic among working-class students at elite institutions, and Lenton’s (2015) economic analysis of the NSS reached compatible conclusions.

I want to hold that account more lightly than the March piece did. It is one plausible mechanism, but it is not the only one, and it risks locating the problem primarily in students’ dispositions rather than in what institutions do with the instrument once it exists. A second mechanism deserves at least equal weight, and it has the advantage of not depending on contested claims about expectation-formation: institutions have every incentive to manage the NSS directly, and considerable latitude to do so, regardless of what any given cohort of students brings with them. This does not require conspiracy or bad faith on anyone’s part; it requires only that institutions learn, over successive cycles, what the metric rewards.

Gaming the instrument: resourcing for the return, not for need

The NSS’s core questions ask about specific, nameable things — academic support, feedback turnaround, communication about wellbeing services, the sense of being listened to. These are exactly the areas in which targeted institutional spending is most legible to the student filling in the survey, and most disconnected from what the instrument is nominally trying to capture, which is the quality of the education itself.

An institution that wants to move its NSS score has a simpler route available than improving teaching across the board: it can concentrate resource on the touchpoints the survey actually asks about. Enhanced personal tutoring in the final year, in which the NSS is administered. Additional bursary and hardship funding, disbursed in ways that generate visible goodwill at exactly the point in the academic cycle when the survey opens. Dedicated wellbeing and academic support staff, whose function is felt directly and immediately by students, in a way that curriculum design or assessment quality is not. The point is not that such spending is improper — often it is necessary and overdue, addressing real and long-standing need. The point is that once a national instrument makes some forms of support more visible than others, institutions have reason to prioritise the visible, independent of whatever their underlying priorities would otherwise have been. None of this is dishonest in the way that manipulating raw data would be dishonest. It is closer to what Goodhart’s Law predicts of any measure pressed into service as a target: once a proxy is known to be watched, behaviour reorganises around the proxy rather than around the thing it was meant to proxy for (Strathern, 1997).

This matters sociologically because it inverts the causal story usually told about NSS scores. A high score is conventionally read as evidence of underlying quality. But if support spending is being allocated, in part, according to its visibility in the NSS return rather than according to demonstrated academic need, then the score can be an artefact of strategic resource allocation timed to the survey window. This is notably true at institutions with a high proportion of bursary-eligible or widening-participation students. There, enhanced support has a double function: it is often genuinely needed, and it is also exceptionally effective at lifting exactly the questions the NSS asks. The same expenditure can be both a good-faith response to student need and a calculated move against a known metric, and there is no way of distinguishing the two from the published data alone.

It would be wrong to suggest none of this is anticipated. The OfS’s response-rate thresholds and post-hoc data quality review exist precisely to guard against interference with the survey process itself: coached responses, inflated participation, and the like. What that apparatus is not built to catch is a subtler and entirely legal practice — reallocating genuine institutional resource toward the specific touchpoints a national instrument happens to measure. Bevan and Hood’s (2006) study of gaming under target regimes in the English public health care system found exactly this asymmetry: audit mechanisms calibrated to catch data manipulation are structurally poor at catching the reshaping of real practice around a known metric, because the latter produces no anomaly for an auditor to find. The safeguard and the exploit are not in tension; they simply operate at different levels, which is precisely why the second survives scrutiny that would catch the first.

Espeland and Sauder’s (2007) work on the reactivity of public rankings is the relevant frame here: once an organisation knows it is being measured and ranked, the measure stops being a passive record of institutional life and becomes a variable that institutions have reason to manage, in ways that could reshape the underlying practice rather than simply reporting on it. Ball (2003) makes a related argument about the “terrors of performativity” in education more broadly — that audit regimes do not just measure institutional behaviour but can produce it, reorganising priorities around what is countable rather than what is valuable. The stakes attached to the NSS are not confined to the survey itself: its results feed directly into the Teaching Excellence Framework, into the major newspaper league tables (the Times, the Guardian, the Mail, among others), and into a wider ecosystem of online ranking platforms that prospective students and their families consult when choosing where to apply. A single set of survey responses, in other words, is asked to carry regulatory, reputational, and market consequences simultaneously. If any metric in higher education were built in a way that invited active management, this is the one.

The rush to publish

This is also, I think, the sociological content behind something anyone who has worked in a university press office will recognise immediately: the scramble to be first out of the gate with results on the day of publication, and the disproportionate weight placed on being “number one” in whatever category can plausibly be claimed — “top in the region,” “first for student voice,” “number one modern university for academic support,” and the like, each true within some sufficiently narrow slice of the data. This need not be read as incidental marketing colour; it could equally be read as the visible tip of the mechanism described above. An institution that has invested in its NSS-relevant provision over the preceding year has every incentive to convert that investment into public recognition as fast as possible, before the news cycle moves on and before any more careful reading of the data — of the kind this piece is attempting — has a chance to circulate. Speed of publication, on this reading, is a strategic choice rather than simply a neutral consequence of having the figures to hand.

A new front: the freedom of expression question

What is genuinely new in the 2026 cycle is the freedom of expression question, asked of students in England only, introduced against the backdrop of the Higher Education (Freedom of Speech) Act 2023. Unlike the support-and-resourcing questions discussed above, this item cannot easily be moved by targeted spending. It asks students to report on something contested and politically live — whether they feel free to express themselves — and is liable to be answered differently depending on a student’s own position in current debates about speech and institutional neutrality, independent of anything the institution has done.

The underlying debate is not manufactured, on either side of it. Concerns about a narrowing of what can be said on campus are substantively held, as are concerns that “free speech” rhetoric can itself be used to shield discriminatory speech from institutional consequence. Both positions have serious proponents and neither is reducible to bad faith. The difficulty is that the NSS was never built to adjudicate between them: it can register a student’s felt sense of freedom, in the aggregate, but nothing about the instrument allows that feeling to be traced back to its cause, whether that is institutional policy, peer culture, national political climate, or something else entirely. Precisely because the cause cannot be isolated, the same movement in the figures — up or down — is available to both sides of the debate as confirmation of whatever they already believed.

The predictable result is that this single question attracts press attention disproportionate to its evidential weight, precisely because it is legible as a scoreboard for an already-polarised debate in a way “assessment and feedback” scores are not.

What to watch for, rather than what the numbers say

• Which institutions publish NSS-based marketing claims within hours of the release, and how does the speed of that response compare across the sector?
• Where support and bursary spending has visibly increased year-on-year, does this correspond to movement specifically on the questions such spending would be expected to influence, rather than on teaching and assessment more broadly? A rolling multi-year comparison against institutions’ published financial statements (via HESA) would be one way to start testing this properly, rather than asserting it from a single year’s figures.
• Does the freedom of expression question receive coverage disproportionate to its methodological weight?

None of these questions are answerable by looking at a single year’s figures in isolation; they are questions about pattern and behaviour over time, not about any one publication day. Nor, it should be said plainly, would any of them straightforwardly prove intent — a sustained correlation between targeted spending and targeted questions is suggestive, not dispositive, and the institutions in question would have every reason to describe it differently. That is rather the point. The problem with the NSS was never simply that it asks imperfect questions. It is that the instrument is well understood enough, by the institutions it measures, that it could be managed rather than merely reported on — and a score produced under active management cannot straightforwardly be read as evidence of the thing it claims to measure.