A college dashboard may display the number of international students beside the number studying part time. Adding the two can look like a quick measure of how many students fall outside a presumed “standard” group. That addition is only valid if the categories do not overlap. A student can belong to both groups, so an unqualified sum can count the same person twice.

The Office for Students' data FAQs emphasise the role of defined populations and measures in interpreting education data. The set calculation below is an independent fictional demonstration. It does not reproduce suppressed cells or estimate any institution's unpublished student characteristics.

Reconstruct the four mutually exclusive cells

Assume an invented institution has 1,000 distinct students in the same reporting population and at the same date. Of these, 300 are international, 250 are part time and 100 are both. “International” and “part time” are stipulated labels here; real publishers may define them differently.

Mutually exclusive group Count
International and part time 100
International and not part time 200
Part time and not international 150
Neither category 550
Total distinct students 1,000

The number who are international or part time, including those who are both, is 300 plus 250 minus 100, or 450. Simply adding 300 and 250 gives 550 and double-counts the shared 100. The complement is 1,000 minus 450, or 550, matching the final non-total row.

Notice that the shared group is one third of the international group but two fifths of the part-time group. These are different conditional percentages because their denominators differ. Neither should be labelled “the overlap rate” without naming the base population.

If the overlap is unknown, keep a range

Suppose only the two marginal counts, 300 and 250, are published. The overlap could be anywhere from zero to 250 under this population total. Therefore the union could range from 300 to 550. Choosing the midpoint would not create a factual estimate; it would merely introduce an unsupported assumption.

In other examples, the total population can force a minimum overlap. If two groups contain 700 and 600 people within a total of 1,000, they must share at least 300. Their union cannot exceed the whole population. This logical bound does not tell the reader which individuals overlap or justify recovering protected information from a suppressed table.

For actual reports, first establish whether rows are mutually exclusive, exhaustive or independently counted. Labels such as “students with any of the following characteristics” often allow multiple membership. A total of percentage rows above 100% can be perfectly legitimate in a multiple-response table, but misleading in a table intended to divide students into exclusive groups.

Reconcile units as well as categories

A person count cannot be added to a course-registration count merely because both columns say “students.” One individual may hold several registrations or attend more than one campus. Check whether the publisher counts people, enrolments, student equivalents or awards before attempting a reconciliation.

Likewise, do not subtract categories from different dates to derive a remainder. An international count from one academic year and a part-time count from another do not share a stable universe. Record the exact reference date and coverage for every term used in an identity.

The final worksheet should show the original categories, known overlap and any unresolved range. When a complete cross-tabulation is unavailable, a bounded answer is more honest than an exact-looking sum. In this invented case, 450 distinct students belong to at least one group because the shared 100 is known; without that information, a precise combined count would be unjustified.