When you see a journal ranking on SCImago (scimagojr.com), the default metric is SJR2 (SCImago Journal Rank 2). Unlike the classic Impact Factor, which simply counts average citations, SJR2 measures prestige – a recursive concept: a journal is prestigious if it is cited by other prestigious journals. This is the same logic that powers Google’s PageRank.

Why prestige instead of raw citations?

A citation from Nature is not the same as a citation from a small regional journal. Prestige indicators give more weight to citations coming from already prestigious sources. The SJR2 method refines this idea by also considering thematic closeness: a citation from a journal in the same narrow field counts more than a citation from a completely different discipline, because field peers are better judges of quality.

Step 1 – Building a weighted citation network

SJR2 uses a 3‑year citation window (e.g., citations from 2008 to documents published in 2005‑2007). Each journal is a node; each citation link is a directed edge. But instead of giving every citation the same weight, two weights are applied:

• Prestige of the citing journal – computed iteratively (see Step 2).
• Thematic closeness between citing and cited journal, measured by the cosine of their co‑citation profiles.

The cosine answers: “Do these two journals tend to be cited together by the same third journals?” If yes, they are thematically related. The calculation uses no predefined subject categories – it emerges from actual citation patterns. This makes the weighting objective and free from arbitrary classification.

Step 2 – Iterative prestige distribution (PageRank‑style)

All journals start with the same initial prestige (1/N). Then, repeatedly, prestige flows from citing journals to cited journals according to the weighted links. Two important safeguards are applied:

• Damping factor (0.9) – a small fraction of prestige is redistributed to all journals to keep the network fully connected (same as Google’s PageRank).
• Dangling node handling – journals with no outgoing references (or very few) do not keep their prestige; it is spread to all journals proportionally to the prestige they normally receive. This prevents artificial inflation.

After the iterative process stabilises (typically 50‑100 iterations), we obtain the Prestige SJR2 (PSJR2) – the total accumulated prestige of each journal. Larger journals naturally have larger PSJR2.

Step 3 – Normalising for journal size (the SJR2 score)

PSJR2 is size‑dependent – a large journal accumulates more total prestige simply because it publishes more papers. To compare journals fairly, we need a size‑independent metric. SJR2 does this by dividing the journal’s share of total prestige by its share of citable documents (articles, reviews, conference papers).

SJR2_i = ( PSJR2_i / (Art_i / ΣArt) )

• SJR2 = 1 → the journal’s prestige per document equals the global average.
• SJR2 = 2.5 → the journal has 2.5 times the average prestige per document.
• SJR2 = 0.7 → 30% below average.

Because the calculation uses shares (fractions that always sum to 1), the weighted average of all SJR2 scores is always 1 every year – no drift over time. This solves a major issue of earlier prestige indicators where scores tended to decrease as more journals were added.

A concrete illustration: two computer science journals

The original SJR2 paper (Guerrero‑Bote & Moya‑Anegón, 2012) provides a striking example (Table 5). ACM Computing Surveys has a 3‑year Impact Factor of 17.97 – very high. Foundations and trends in communications and information theory has an Impact Factor of 7.20. Their SJR2 scores, however, are 2.84 and 12.06 respectively – the second journal ranks much higher. Why?

• The second journal receives citations from journals with much higher thematic closeness (cosine 0.67 vs 0.13).
• The citing journals themselves are more prestigious (average SJR2 of citations: 2.07 vs 0.92).
• The first journal publishes many more citable documents (361 vs 10), diluting its prestige per paper.

SJR2 thus rewards focused, specialised journals that are highly valued by peers in the same niche – even if their raw citation counts are modest.

How SJR2 compares to other metrics

The SJR2 method was tested against a 3‑year Impact Factor and the SNIP indicator (field‑normalised citation impact). Key empirical findings from the original paper:

• High correlation overall (Spearman > 0.94 with JIF) – SJR2 is not radically different, but refines the ranking.
• Better field equalisation than JIF – The distribution of SJR2 scores is more balanced across subject areas, especially inside fine‑grained specific subject categories (295 of them).
• Cosine weighting increases intra‑field prestige flow – Journals receive a larger share of prestige from journals within the same specific subject area (from about 34% of raw citations to 52% of weighted prestige).

Why the SJR2 method matters for research evaluation

For university administrators, funding agencies, and researchers, SJR2 provides a field‑normalised, size‑independent, stable measure of journal prestige. It does not penalise journals in low‑citation fields (like mathematics or social sciences) because the cosine weighting gives full credit to citations from thematically close, high‑prestige sources. And because the global average is always 1, you can track a journal’s relative performance over years without worrying about inflation or deflation of scores.

Today, SJR2 is the default ranking metric on the SCImago Journal & Country Rank portal (scimagojr.com), covering over 30,000 journals from Scopus. When you see a journal’s SJR2 quartile (Q1, Q2, etc.), it tells you how that journal’s prestige per paper compares to others in its category – a direct, interpretable indicator of standing.