Schools below the SALSA prediction

Key findings

SALSA is a statistical model from Skolverket that adjusts school results for pupil composition. For each school it produces an expected merit value based on parents' education level, the share of newly arrived students, and gender distribution. The residual — actual merit minus expected merit — tells you whether the school performs better or worse than the composition predicts, regardless of how high or low the absolute level is.

Schools with a residual below −10 are those that most noticeably deviate from the model's prediction. This is not a quality verdict on the school.Several schools on the list work explicitly with students who require extensive support, which SALSA does not see.

Schools with large SALSA deviations — read each school in context

The table shows schools sorted by SALSA residual (most negative at the top). The residual is shown with its 95 % confidence interval based on the year-9 cohort size; rows where the interval includes −10 are de-emphasised and should be read with extra care.

Showing top 50 of 370 ranked schools with SALSA residual ≤ −10.17 independent and 33 municipal among these. An additional 63 schools matched the threshold but are excluded from the ranking (see below). Click a school name for its full profile with context.

Schools excluded from the ranking

Comparison: the most successful schools

For contrast, the 5 schools with the most positive SALSA residual are shown — schools that performbetter than what the student composition predicts.

Methodology and limitations

The SALSA residual is calculated per school as actual merit value minus model-predicted merit value. According to Skolverket's technical documentation the model explains roughly half of the variation in merit values between schools (the exact figure varies by cohort) — the remaining variance depends on factors such as leadership, resources, pedagogical quality, and individual student support needs, which the model does not measure.

SALSA's most important blind spot is that it does not account for students' individual support needs. Schools that take in children with autism or ADHD diagnoses, severe language disorder, or other extensive support needs — so called resource or intervention schools — will by definition show negative residuals because the model does not know the students have elevated needs. These schools may do exceptional work with their students without it being visible in SALSA.

To handle this blind spot, the following schools are filtered out of the ranking and shown in a separate table:

• The school is an adapted compulsory school, adapted upper-secondary school, special school, or Sámi school — where merit value is not an appropriate outcome measure.
• The organiser (huvudman) is on Skolkoll's curated list of organisations whose primary business is operating resource or intervention schools (Specialpedagogiska skolmyndigheten, Utvecklingspedagogik Sverige AB, Nytida, Adventum, Aktivia, and others). This list is Skolkoll's classification — not a government register — and is reviewed periodically.
• The business description in Bolagsverket's register explicitly states resource or intervention provision, targeting pupils with NPF, severe language disorder, autism or ADHD, or a stated special-education focus.
• The year-9 cohort is below 15 pupils, which is Skolverket's own publication floor for SALSA. Residuals on very small cohorts are statistically unstable, and Skolverket suppresses them precisely to protect the privacy of small pupil groups.

For the ranked schools, the residual is shown with an estimated 95 % confidence interval (Z · σ / √N₉, with σ ≈ 75 as Skolkoll's approximation for the standard deviation of merit value — the national spread varies between roughly 70 and 80 across cohorts). The interval describes the uncertainty in the school's actual merit mean; when it includes the −10 threshold (de-emphasised row) the signal is too uncertain to support individual conclusions about the school.

The threshold of −10 residual points is an analytical choice. Lower thresholds (e.g. −5) yield more schools but a weaker signal; higher thresholds (e.g. −20) yield fewer but clearer deviations.

General supervision picture for Swedish compulsory schools

Source: Swedish Schools Inspectorate planned supervision, latest available year. The percentages refer to the proportion of inspected schools that received at least one decision with deficiencies.