What is the sample size, methodology, and margin of error of the poll projecting Scottish Labour winning 13 MSP seats?

Polls like the recent More in Common survey, which projects Scottish Labour winning just 13 seats in the Scottish Parliament, use a method called Multi-level Regression and Post-stratification, or MRP. This technique builds detailed predictions for each constituency by combining a survey of voting intentions with data on demographics like age and past voting patterns, rather than relying on simple national averages. Exact sample size and margin of error for this poll aren't clearly stated in the released results, which makes it harder to judge how reliable the numbers are. In contrast, a comparable Scottish survey by Find Out Now used responses from 4,105 adults collected online in March 2026 and weighted them to match population figures for gender, age and social grade. Standard guidance suggests a sample of about 1,000 people usually produces a margin of error around 3 percent for national vote shares, though MRP can narrow some uncertainty by borrowing patterns across groups while adding its own assumptions.

This matters to teenagers because polls influence how parties campaign and what issues they prioritise, from education funding and mental health support to climate policies that will shape your future jobs and environment. Supporters of MRP argue it gives sharper constituency-level forecasts than traditional polls, helping voters understand likely outcomes under Scotland's Additional Member System, which mixes constituency and regional list seats. Critics point out that without full details on sample size or how non-sampling errors like turnout shifts are handled, the projections can overstate certainty and affect public trust in the process. Checking multiple sources and understanding these limits helps young people form their own views rather than taking any single headline at face value.

The More in Common poll projecting Scottish Labour to win just 13 MSP seats under the Additional Member System relies on multi-level regression and post-stratification (MRP), a modelling technique that estimates constituency-level outcomes by combining demographic predictors with national survey data rather than raw aggregates. Unlike conventional constituency polling, MRP borrows statistical strength across subgroups to generate seat projections, yet this approach introduces model-based assumptions that can amplify uncertainty when constituency-level predictors are weak. Precise sample size and margin of error for the More in Common exercise remain undisclosed in the PA Media release, which instead highlights regional patterns such as potential Conservative advances in Banffshire and Buchan Coast. This absence of technical detail contrasts with more transparent reporting elsewhere.

A comparable Scottish MRP conducted by Find Out Now for Electoral Calculus surveyed 4,105 adults online between 13 and 31 March 2026, weighting responses for gender, age, social grade and recalled past vote to align with Office for National Statistics benchmarks. Ballot Box Scotland guidance indicates that a 1,000-respondent sample typically produces a margin of error around three percentage points for national vote shares; MRP can narrow effective intervals by pooling information across areas, but non-sampling errors—turnout misestimation, late swings and panel recruitment bias—often widen credible ranges. Ipsos analysis of recent UK elections suggests roughly a nine-in-ten probability that true vote shares lie within broader bands than classical margins imply. Opinium’s Scottish weighting similarly adapts UK-wide practices, yet remains sensitive to the quality of historical constituency data.

These methodological choices intersect with emerging policy debates over mandatory disclosure of MRP specifications and minimum sample-size thresholds for Scottish-specific forecasts. Proponents argue that standardised reporting would enhance democratic accountability by allowing voters and parties to assess projection reliability before Holyrood elections. Critics caution that rigid standards risk stifling methodological innovation and may prove difficult to enforce across commercial pollsters. Empirical evidence from the 2026 cycle, where Labour’s projected constituency losses forced reliance on regional lists for all 13 seats, underscores how small shifts in modelled list-vote shares can produce disproportionate seat changes under the Additional Member System. Cross-referencing GB-wide polls with Scottish subsamples further reveals persistent challenges in achieving representative online panels, particularly for smaller parties. Balancing transparency requirements against practical constraints therefore remains central to improving the informational value of electoral projections without unduly constraining statistical flexibility.