How to Find the Original Source of a Statistic

When it is repeated often enough, the embodiment of a number achieves its own weight. A figure that sounds authoritative when it is cited. But the rule to follow is: a statistic should not be repeated until it is traced back to the original report, study, or dataset, and the first task is to identify exactly what kind of claim it is — a statistic, a scientific finding, or a quote/testimony. That is because the figure could have changed between the time it was first measured and the time it is quoted, or the precise question it was meant to answer could be unclear.

To trace the statistic to its origin and verify its accuracy, the first step is to isolate the claim and identifying its type. Next, they follow the citations backward, using the exact wording or distinctive phrasing to locate the earliest repetition of the number. Searching the number itself with a discipline keyword can often surface the underlying study or report directly. Combining the number with an identifiable organization can also lead to the source or the statement closest to it.

Engineers should expect that the "source" may not be a journal article, but a report or data release from a statistical agency. For example, economic and public-health statistics often come from national statistical agencies rather than journal articles, and the original "thing to be cited" is an official report or data release.

Assessing the credibility of a statistic also requires identifying the originator of the statistic, rather than the publisher. CSSLibraryGuides states that in Statistical Abstract, the original source of the statistics is listed at the bottom of the table or at the bottom of a range of related tables.

Most important, though, is to identify "published by whom, in what year and how directly"? The publication date matters because a statistic may be outdated and superseded by newer research or data. It may be accurate, but no longer true, if the underlying population, question, or boundaries have changed. Verification requires a clear idea of what it is that is being measured: it can be argued that that if the number is treated as a picture, then the three dimensions are the method, the scope, and the time.

That is why the verification habit should be to record the statistic, the primary source organization, the publication year, and a direct URL in a searchable reference list. Using library and database clues, it is often possible to trace the number directly to the original statistical source working back through abstracts, tables, and source boxes. At the top of the most credible sources, all of the source information is attached to the data.

However, many sources include statistics in visually appealing formats with related measurements in different quantities, that are not all traced back to the same survey or dataset. If the end user has a copy of the data in their hands, that is the closest verification. But relying on one source to represent a number can be dangerous, especially in health and economics. Triggers for further verification should be the organization's declared emphasis, the presence of outdated by more than a year, and changes in the number between the source document and the quoting text.