← Back to Blog

How the Integrity Score Works, and What It Does Not Measure

Integrity Index editorial team

Every scored member on Integrity Index carries a number from 0 to 100 and a letter grade. This article explains exactly how that number is built, using the same rules the scoring code runs, and is just as direct about the things the score cannot tell you.

The short version: the score summarises how much of a member's public record is available and how that record looks on a set of transparency and accountability signals. It is not a verdict on anyone's honesty, and it does not rate their politics.

Twelve factors, each scored from 0 to 1

The score draws on up to twelve factors. Each one is calculated separately from a specific public data source and produces a value between 0 and 1. Each factor also has a nominal weight:

FactorWhat it looks atWeight
Disclosure and conflictFiled disclosures, conflict flags, unsourced or placeholder entries, concentration of holdings in one sector0.20
Vote participationShare of recorded divisions the member took part in0.12
Lobbying exposureRegistered lobbying communications with the member0.12
Procurement proximityGovernment contracts linked to the member0.12
Trading activityReported securities trades0.12
Political financeContribution records0.12
Expenditure prudenceOffice and travel spending reports0.12
Committee proximityCommittee memberships and the sectors they oversee0.10
Committee oversightCurrent committee roles, including chairs0.10
Travel transparencyDelegation travel and whether it was publicly disclosed0.08
Legislative activityBills sponsored, used for provincial legislatures0.08
Data provenanceHow fresh and complete the underlying records are0.04

The weights are nominal because no member has all twelve factors. Federal MPs, for example, have no securities-trade stream, and most provinces publish no lobbying registry in a form we can ingest. So the score is a weighted average over only the factors that exist for that member. If a member has five factors, those five weights are rescaled to add up to one. A missing factor neither helps nor hurts.

Two adjustments after the average

Confidence. A score built from three data streams is less certain than one built from eight. Members with exactly three data factors have their average multiplied by 0.95; members with four or more are not reduced. (Data provenance does not count as a stream here, because it describes the other data rather than the member.)

Cross-stream overlaps. Some patterns only show up when two sources are read together. When the organisations or subjects in a member's lobbying contacts overlap substantially with the sectors in their disclosures or trades, or when industry-tagged contributions line up with disclosed holdings in the same industry, a small multiplier (0.95 to 0.97 per overlap) applies. The combined effect is capped: these multipliers can never take a score below 75 percent of its pre-adjustment value. An overlap is a reason to look more closely, not evidence of wrongdoing.

How the disclosure factor works

Disclosure carries the most weight, so its rules are worth spelling out. A member starts at 1.0. Each disclosure that the conflict review flags removes 0.10. Entries that are generic placeholders or lack a source link reduce the score by at most 0.10 each, in proportion to their share of the member's filings, so members who disclose a lot are not penalised for the volume. If more than 70 percent of a member's holdings sit in a single sector, a small concentration deduction applies. The result is then blended with a measure of how broad and substantive the holdings are, and reduced further when flagged assets make up a large share of the disclosed dollar value.

How vote participation works

Attendance is measured against the chamber, not against whatever ballots we happen to hold. A member's participation is the number of recorded divisions they took part in, divided by the number of divisions held in their current session. A yea, nay or recorded abstention counts as taking part. "Didn't vote" and "paired" do not.

Pairing is a formal arrangement in which two members on opposite sides agree that neither will vote, often so that one of them can be away on government or constituency business. It is a legitimate practice. We still count it as non-participation because the score measures presence in the record, not the reason for an absence. That is one reason the grade should never be read on its own.

The chamber-wide denominator only applies once we hold at least 20 ballots for the member. Below that, we divide by the ballots on file instead, so a partially loaded record does not register as mass absenteeism.

From number to letter

ScoreGrade
90 and aboveA+
85 to 89.99A
80 to 84.99B+
75 to 79.99B
65 to 74.99C
55 to 64.99D
Below 55F

What the federal distribution looks like

As of September 2026, 333 of the 346 sitting members of the House of Commons have a score. They range from 60.6 to 92.4, with an average of 75.6. By grade:

GradeMPs
A+3
A21
B+35
B98
C168
D8
F0

C is the most common federal grade, and that is by design. The scale is not curved so that the average member earns a B. A C means the public record has some gaps or signals worth reading, usually a mix of missed divisions, flagged or thinly sourced disclosure entries, and lobbying or spending that the factors pick up. Open any profile's score breakdown to see which factors pulled that member's number up or down, and by how much.

When a member gets no grade

We would rather show no grade than a misleading one. A member is left unscored when:

  • they are a federal MP with no public disclosure on file, since disclosure is the backbone of the federal score;
  • they sit in a legislature whose disclosure filings we cannot yet ingest (currently Manitoba, Newfoundland and Labrador, and Prince Edward Island);
  • fewer than three independent data streams exist for them.

An unscored member is not a failing member. The profile says "Unscored" and gives the reason. Our coverage article lists where each legislature stands.

What the score does not measure

A single number cannot capture a legislator's work, and this one does not try to. It does not reflect:

  • Positions or ideology. How a member votes is reported on their profile but never scored. Voting against a bill is not penalised; only not voting counts.
  • Effectiveness. Constituency casework, committee reports, speeches and behind-the-scenes negotiation leave little trace in the data we use.
  • Legitimate reasons for absence. Illness, parental leave, ministerial travel and pairing all lower participation.
  • Private conduct. The score only uses public records. It says nothing about matters that were never disclosed or never investigated.

Our conflict signals are automated matches between public datasets. A flag means two records line up in a way a reader may want to check, not that an ethics rule was broken. Findings of actual contraventions come from the federal Conflict of Interest and Ethics Commissioner and the provincial integrity commissioners, and every profile links back to the official sources so you can check them yourself.

How to use it well

Treat the grade as a starting point. Open the methodology for the data sources, compare two members side by side on Compare, and read the factor breakdown before drawing conclusions. If you find a record that looks wrong, tell us. Corrections to the underlying data flow into the next score recalculation.