Methodology
Every report ships its own methodology block, and every figure traces to a named source. The standards below govern all of them.
Sources, by hierarchy
Every figure traces to a named source the reader can check, drawn in priority order: primary and official records first, then peer-reviewed research, then established monitoring organizations. Each source is labeled by type, so a measured share is never presented as the same kind of evidence as an organizational tally or an advocacy estimate.
First-party measurement
Where the Institute measures directly, it shows the method. A search-attention tally tracks US Google Trends interest for curated terms, month by month, reported as year-over-year change; corpus studies code a defined sample, for example newspaper headlines, against a fixed codebook, with inter-rater reliability reported. Each metric measures one specific thing, attention or a coded share or a volume, never belief or behavior.
Correlation, not causation
Where online activity relates to offline events, reports write “coincided with” or “rose alongside,” never “caused.” Several cited authors make no causal claim, and that posture is preserved.
Aggregates only
Every number is a share, count, rate, or model score, never a claim about an individual. The Institute does not profile or name a person, and aggregates use minimum-cell suppression so no individual is identifiable.
Uncertainty, disclosed
Preprint versus peer-reviewed status, sample sizes, lower-bound flags, and contested framings are carried into the prose. Where a label is model-inferred it is treated as probabilistic, and a low-confidence attribute is shown as “undetermined,” never guessed.
Contested questions
On a contested subject, for example whether opposition to Zionism is antisemitism, or the framework used to code it, the Institute measures the discourse and reports the competing findings by name. It does not adjudicate the claim itself.