Election Intelligence Methodology

All election intelligence features rely exclusively on official, publicly available data sources. Primary sources include the European Parliament plenary vote records, the EU Transparency Register for lobby meeting data, Eurobarometer surveys for citizen sentiment alignment, and the official MEP activity records maintained by Parliament.

National election data is sourced from official election authorities in each member state. We do not use polling data, media reports, or unofficial sources for any metrics.

Data is ingested through automated pipelines that pull from official APIs and structured data exports. All raw data is preserved in our database for auditability.

Plenary votes are categorized by policy area using a combination of the European Parliament's own committee assignment, the procedure's legal basis, and keyword analysis of the legislative text title and summary.

Each vote is assigned to one or more of the following categories: Digital & Technology, Environment & Climate, Economy & Trade, Migration & Justice, Defence & Security, and Social Policy. When a vote spans multiple categories, it appears in all relevant categories.

For the Party Voting Compass, a party group's position on a given vote is determined by the majority behavior of its members. If more than 60% of the group's participating MEPs voted the same way, the group is classified as "For" or "Against". If participation falls between 40% and 60%, the group is classified as "Mixed".

Attendance is calculated as the percentage of plenary roll-call votes in which the MEP participated (voted for, against, or abstained) out of the total number of roll-call votes held during the measurement period.

Absences due to officially documented medical leave, parental leave, or official delegation travel are excluded from both the numerator and denominator. Committee meetings and other non-plenary activities are not factored into this metric.

The chamber average is computed across all sitting MEPs for the same period, providing a benchmark for comparison. Attendance is reported as a simple percentage rounded to the nearest whole number.

Party loyalty score measures the percentage of roll-call votes in which an MEP voted in alignment with the majority position of their party group. The majority position is defined as the direction in which more than 50% of the group's participating members voted.

Only votes where the group had a clear majority position (more than 50% alignment) are included in the calculation. Free votes designated by the group, procedural votes, and votes where the group was evenly split are excluded.

This metric is descriptive, not evaluative. A high loyalty score may indicate strong party cohesion or it may indicate lack of independent judgment. A low loyalty score may indicate principled dissent or it may indicate disengagement. GovLens does not assign qualitative meaning to any score.

Citizen sentiment alignment measures the degree to which an MEP's voting record corresponds to public opinion as measured by Eurobarometer surveys. This is an experimental metric.

For each policy area where both a plenary vote and a corresponding Eurobarometer question exist, we compare the MEP's vote direction with the majority public opinion in their home country. The alignment score is the percentage of matched comparisons.

This metric has significant limitations: Eurobarometer questions may not perfectly correspond to the specific legislation voted on, public opinion is measured at a different time than the vote, and national-level public opinion may not reflect the MEP's specific constituency. These limitations are disclosed alongside every alignment score.

The promise tracker (planned feature) will compare stated campaign positions with actual voting behavior. Promises are sourced exclusively from official party manifestos and candidate platforms published during the election campaign period.

A promise is considered "kept" if the MEP voted consistently with the stated position on all relevant roll-call votes. A promise is "broken" if the MEP voted against the stated position on the majority of relevant votes. A promise is "in progress" if relevant legislation has not yet reached a final vote.

Promises that are vague or unmeasurable (e.g., "fight for a better Europe") are excluded from tracking. Only specific, verifiable commitments tied to legislative outcomes are tracked.

GovLens is committed to non-partisan, data-only election intelligence. We do not rate, rank, or recommend candidates or parties. We do not use value-laden language such as "good attendance" or "poor loyalty". All metrics are presented as raw data with contextual benchmarks.

We do not publish aggregate scores that combine multiple metrics into a single "performance" number. Combining attendance, loyalty, and other metrics into a single score would require subjective weighting decisions that we believe are for voters, not platforms, to make.

Color coding is used strictly for categorical distinction (for/against/mixed, above/below average), never to imply quality judgments. Green and red indicate direction relative to average, not "good" and "bad".

All data includes clear source attribution. Users can verify every figure by following source links to the original official data. We publish our methodology openly and welcome corrections.

If an MEP, party, or member of the public believes any data point is incorrect, they can submit a correction request through our feedback system. Correction requests are reviewed within 5 business days.

For factual errors (incorrect vote count, wrong date, misattributed data), we issue corrections immediately upon verification and log the correction publicly in our changelog.

For methodological disagreements (e.g., how a vote should be categorized), we review the concern and, if warranted, update our methodology. Methodology changes are documented with a rationale and applied retroactively to ensure consistency.

We maintain a public corrections log documenting all changes to published data, including the date, nature of the error, source of the correction, and the corrected value.