The Accountability Gap

Every MP generates three signals. They vote in Parliament — that is the "do" signal. They speak publicly about policy — that is the "say" signal. Their constituents sign petitions on the issues they care about — that is the "want" signal.

Polis measures the distance between these three signals for 639 MPs. When what an MP does, says, and what their constituents want all point in the same direction, the accountability score is low. When they diverge — when an MP votes on everything but says nothing, or when Parliament's agenda bears no resemblance to local priorities — the score climbs.

The results reveal something structural about how British democracy functions under a large parliamentary majority.

One MP

Bethnal Green and Stepney. A constituency in east London where one in five residents lives in overcrowded housing, NHS waiting times rank in the worst quartile nationally, and food bank referrals have risen for three consecutive years. Service pressure: 71 out of 100.

Rushanara Ali has voted in hundreds of divisions since the 2024 election. Her voting record spans education, justice, housing, health, the economy, and immigration. In almost every domain, her "do" score — the proportion of her parliamentary activity allocated to that topic — exceeds 60.

Her "say" score — how much she talks publicly about those same topics, measured through media appearances and extracted claims — is functionally zero across every single domain.

Her constituents' "want" score — how intensely they petition on each topic — is in single digits for most areas, peaking at immigration and animal welfare.

Three signals. None of them aligned. That is what the accountability score measures.

The Pattern

When you plot every MP's accountability score against their constituency's service pressure, the distribution tells an uncomfortable story. The bulk of Parliament — 366 of 639 MPs, or 57% — scores between 50 and 70. Only 38 MPs (6%) score below 20.

The national average accountability score is 50. More than half of all MPs have a measurable gap between what they do in Parliament, what they say in public, and what their constituents express through petitions.

The accountability score does not measure effort. It measures alignment — between parliamentary action, public communication, and constituent priority. Most MPs fail this test.

The Funnel

Look at how the numbers narrow.

639 MPs have sufficient data for an accountability score. 366 score above 50. 36 score above 70 — a severe gap between their parliamentary activity, their public statements, and their constituents' expressed priorities.

Every single one of those 36 MPs is Labour or Labour (Co-op).

This is not a partisan attack. It is a structural observation, and understanding why it happens is more instructive than the headline.

The Machine

The 2024 general election delivered Labour a 170-seat majority. In parliamentary terms, this means the government can whip its backbenchers through virtually any division and win. The result is that Labour MPs vote at extremely high rates across every policy domain — education, health, housing, justice, foreign affairs, the economy.

This inflates their "do" score to 70, 80, even 90+ across multiple domains. It does not mean they are personally invested in each topic. It means the whip told them to be in the lobby.

Meanwhile, their "say" score — derived from media appearances, public statements, and extracted claims — sits near zero. Of 1,065 claims extracted from parliamentary media, they are unevenly distributed. Most Labour backbenchers have generated no trackable public claims at all.

The accountability score computes the total variation distance between three domain distributions: what they do, what they say, and what their constituents want. When the "do" distribution is flat and high (voting on everything equally), the "say" distribution is empty, and the "want" distribution is concentrated on two or three issues, the resulting gap is large.

Conservative MPs average 43.9 — lower than Labour, but not because they are more aligned with their constituents. They simply vote less often as the opposition, which produces a smaller "do" signal relative to "say" and "want". The Lib Dems sit between the two at 49.3.

The SNP averages 23.7, the lowest of any major grouping. This is not because Scottish National Party MPs are unusually responsive to their constituents. It is because they frequently abstain from Westminster divisions on matters they consider devolved to Holyrood, which compresses their "do" distribution.

The score is revealing something real — but what it reveals is structural, not personal.

The Contrast

The accountability score's most instructive feature is what happens when you compare two MPs from the same party.

Both represent London constituencies with near-identical service pressure. Both are Labour. But David Lammy's accountability score is 16.3 and Rushanara Ali's is 70.7.

The difference is ministerial. Lammy is Foreign Secretary. He votes far less frequently (ministerial duties prevent regular lobby attendance), which reduces his "do" score. But he generates substantial media coverage, which populates his "say" score. The result: a lower total variation distance between his three signals.

Ali, by contrast, votes at full rate (whipped through every division), generates minimal trackable public claims, and represents a constituency whose petition priorities diverge from the legislative agenda. The gap widens in every direction.

This comparison reveals a genuine limitation. The score rewards ministers for voting less and having media profiles. It punishes backbenchers for doing what their party asks. Whether this reflects real accountability or a measurement artefact depends on what you believe accountability means.

The Stress Test

If the accountability score rewards non-voting, then the Prime Minister — who votes least of all — should score absurdly well. He does.

Keir Starmer scores 6.1 — the best accountability score in Parliament. Better than the Speaker (7.5), who barely votes by convention. Better than any backbencher in any party.

This is the score doing exactly what it says: measuring the distance between three signals. When the "do" signal is nearly silent (the PM votes in a handful of divisions per session), the distance to "say" and "want" is necessarily small.

But his reality gap is 53. His service pressure is 57 — nearly double the national average. The community he represents faces genuine material challenges, and the domain distribution of parliamentary activity bears little resemblance to those challenges.

The accountability score and the reality gap are measuring different failures. One measures alignment between an MP's actions and their community. The other measures alignment between Parliament's agenda and local need. Starmer fails the second test while acing the first.

The Do-Want Chasm

Underneath the per-MP scores lies a topic-level pattern that no amount of structural explanation can account for.

Foreign affairs has a 92-point gap. MPs vote on it constantly; constituents almost never petition about it. Education sits at 71 points. Justice at 70. Housing at 68.

These are not fringe issues. Housing, education, and health are the three domains where constituency-level service pressure is most concentrated. They are also the three domains where the "want" signal — petition signatures — most clearly reflects material conditions.

Parliament votes on them. Constituents petition about them. But not at the same rate, not in the same proportion, and not with the same emphasis. The do-want gap measures this structural misalignment across the entire House of Commons.

What This Means

The accountability score exposes structure, not character. The 36 MPs scoring above 70 are not necessarily worse representatives than those scoring below 20. They are backbenchers in a governing party with a massive majority, subject to a whip that demands their presence in the lobbies on every major division. The score measures the consequences of that structure.

The "say" signal is catastrophically weak. Of 639 MPs, the overwhelming majority generate no trackable public claims about policy domains. 1,065 claims from 93 MPs means most of Parliament is publicly silent on the topics they vote on. This is not a data gap — it is a transparency gap. MPs who vote on education, housing, and health policy but never explain their positions publicly are failing a basic accountability test regardless of their score.

The do-want chasm is the real finding. Party structure explains why Labour backbenchers score high. It does not explain why education has a 71-point gap between parliamentary activity and constituent petitions. That gap reflects a disconnect between what Parliament prioritises and what communities experience — and it persists regardless of which party governs.

Cabinet ministers game the metric unintentionally. The PM and senior ministers score low because they vote infrequently, not because they are well-aligned. This is a known limitation. The reality gap provides the corrective: Starmer's 6.1 accountability score and 53.0 reality gap, read together, paint an accurate picture.

What's Missing

The "say" signal is the weakest link in this analysis. It relies on media appearances and extracted public claims — 1,065 in total, from 93 MPs. The remaining 546 MPs have no captured claims. A more comprehensive dataset (constituency surgeries, social media posts, local press interviews) would strengthen the "say" distribution substantially.

Petition data captures a specific kind of engagement — digitally literate, motivated enough to sign, and concentrated on topics where petitions exist. It does not capture views expressed at community meetings, in letters to MPs, or through local campaigning. The "want" signal is real but partial.

The score treats all policy domains equally. A 70-point gap in foreign affairs may be less consequential for constituents than a 50-point gap in housing. Domain weighting by local relevance would improve the signal but introduces editorial judgment into what should be a mechanical calculation.

Finally, the score does not account for time. An MP who has been vocal on housing for three years but quiet in the past six months will have a "say" score that underweights their recent silence. Temporal decay is not yet implemented.

How We Measured It