When does a long historical comparison tell a different story?

An old destination and a recent acceptance are not two points on one line

The older Eton example describes final leaver destinations. The later count’s published basis identifies apply-centre acceptances. Even if both are labelled Oxford plus Cambridge, joining them would change the outcome being measured underneath the line.

The useful historical question is which like-for-like series is held within each period. Preserve the destination archive, examine the later admissions series on its own terms, and leave the bridge unresolved unless compatible evidence supports it. A coincident overseas increase would not by itself prove that it caused a domestic change.

Count the years that actually contribute

A selected calendar window can contain missing records, incompatible qualifications or conflicting publications. Matched years are the dates on which the selected schools have eligible evidence on the same definition. They need not be consecutive, and two apparently equal windows may contain different numbers of usable observations.

A missing year is not a zero and should not be filled by a nearby year or an approximate period. Multi-year summaries remain summaries of their published period; they do not create an annual observation for each year inside it.

Test the question without moving the goalposts

The comparison reports matched annual differences and uses the existing period checks where enough comparable evidence is available. Those checks examine all matched years, the latest three and five matched years, and leaving out each matched year. They do not test every possible window. The claim checker supplies the separate broader window analysis.

If definitions change within the requested period, the reviewed model splits the matched evidence into constant-definition segments. A single median across different definitions is not manufactured to provide a tidier conclusion. “Insufficient comparable evidence” is a result about the question’s support, not a verdict on either school.

Treat the endpoint as a new question

After finding a pattern, ask whether it appears on a broader grade band, whether the latest observation is provisional, and whether application volume or cohort scope also changed. These are specific follow-on questions. A stable descriptive pattern still does not identify teaching quality or the cause of the difference.

Supporting evidence and limits

Open a record to inspect its full definition, publication references and the scope of any checks. A source reference alone does not verify a figure.

1. Eton College · 2015 · Leaver destinations · 2000–2015 · 2015

   The earlier source uses a final leaver-destination basis.

   No adequate source description or original-publication check is presently held in this public record. Any reference keys are retained for further research.

2. Eton College · 2025 · Oxbridge destination counts · later years · 2025

   The field’s published basis identifies apply-centre acceptances; not substituted for the earlier destination measure.

   No adequate source description or original-publication check is presently held in this public record. Any reference keys are retained for further research.

• The two destination-related examples are deliberately not subtracted or plotted as one trend.
• A lack of statistical significance is not proof of institutional resilience; no regression results from the older dossier are imported.