# Scotland Counted council-performance index: methodology

## What the score answers

The headline score asks: **which Scottish councils show the broadest measurable strength in resident experience, service outcomes, council management, value for money and their own operational climate record?** It does not claim to measure every reason a place is good to live, and it does not treat poverty, crime, health or affluence as proof that the council itself is good or bad.

The July 2026 Local Government Benchmarking Framework contains 109 indicators. This version uses **53 distinct official LGBF measures**, capped into **29 evidence families**. The remaining measures are either context, ambiguous inputs, stale, not applicable to all councils or too easy to misread as performance.

## Weighting

| Pillar | Weight |
|---|---:|
| Resident experience | 25% |
| Service outcomes | 40% |
| Council management | 20% |
| Value for money | 10% |
| Operational climate | 5% |

Indicators are not allowed to vote independently. They are averaged within evidence families; families are averaged within domains; domains are averaged within pillars. This stops education or satisfaction from dominating simply because more questions happen to be published for them.

Within service outcomes, children and schools receive 40% of the pillar, environment and roads 30%, and adult care 30%. The other domains are equal within their pillar.

## Normalisation and direction

The public score converts each measure to a 0-100 percentile across all 32 councils. The Improvement Service direction is used for higher-is-better and lower-is-better measures. The gender pay gap is scored by distance from zero. Spending outturn is scored by distance from 100% of budget.

Missing or suppressed values receive a neutral score of 50 for that one indicator and reduce the published coverage figure. They are not treated as zero. Modelled source values remain in the central estimate, are flagged, and are neutralised in half of the robustness simulations.

## Why costs are paired with outcomes

Low spending is not automatically good. Every cost in the value-for-money pillar is paired with a quality, outcome or resident-satisfaction measure. A geometric mean penalises both cheap-but-poor services and strong-but-very-expensive services. Costs still receive only 10% of the headline score because rurality, geography, demand and accounting practice affect them.

## National and peer views

The national score answers what residents actually observe across Scotland. A second score repeats the calculation inside the Improvement Service's official family groups of comparable councils. The peer score is a diagnostic, not a replacement: adjusting away deprivation or rurality can hide real differences residents experience, while ignoring them can unfairly punish councils with harder conditions.

## Robustness

The analysis runs 10,000 deterministic simulations. They vary percentile, winsorised min-max and winsorised z-score normalisation; arithmetic and geometric aggregation; pillar, domain and family weights; whether shared hospital-coordination outcomes are included; and whether known modelled values are neutralised. The public output reports median rank, a 5th-95th percentile rank interval and the probability of appearing in the top five or top ten.

These probabilities describe stability across plausible model choices. They are not probabilities that a council is objectively "the best".

## Kept separate from the headline score

- child poverty, earnings, claimant rates, GVA and other labour-market conditions;
- recorded crime, life expectancy, drug deaths and wider health outcomes;
- overall neighbourhood rating, loneliness and housing condition;
- Crisis Grant acceptance and processing, because applicant mix and eligibility interpretation are not observed;
- council-tax bills and budget gaps, because price and financial pressure are not service quality;
- current accountability-record counts, because research depth and audit cycles differ by council;
- council-landlord measures, because six councils do not own comparable housing stock.

These appear as standalone context or accountability views so readers can inspect them without confusing circumstance with council competence.

## Reporting periods

The index is a latest-comparable snapshot, not a single-year table. Most measures are 2024-25. Complete adult-care experience and road-condition data use 2025-26. Some satisfaction measures are three-year rolling estimates. Every underlying row retains its reporting period.

## Important limits

The Local Government Benchmarking Framework is designed as a diagnostic "can opener", not an official league table. Council services operate alongside the NHS, schools, households, employers and national government. Survey estimates have sampling error; council accounting and case mix vary; small numbers are sometimes suppressed. The score is an editorial model whose assumptions are published and stress-tested, not an official judgement.