A fictional cohort with different starting points
Imagine a program enrolling both early-career students and experienced managers. Its graduates' salaries reflect different prior experience, occupations, hours, and locations. A single median blends those paths.
An applicant entering from a different background may not resemble the middle graduate. That does not make the statistic useless; it means the relevant subgroup and context matter.
The missing comparison
If graduates earn more after the program, some change might reflect experience, a new employer, broader labor-market conditions, or selection into the degree. The before-and-after difference does not by itself estimate what would have happened without enrollment.
The NIST statistical handbook explains descriptive summaries such as the median. A descriptive statistic can be calculated correctly while being asked to answer a causal question it was not designed to resolve.
Read the footnote before the headline
Useful details include: - The graduating cohort and time since graduation. - The number responding and the number eligible. - Whether nonemployed graduates are included in any employment measure. - Whether salary means base pay, total compensation, or another quantity.
Salary among employed respondents and earnings across all graduates are different summaries. A report should make that distinction visible.
Use scenarios instead of one promised number
A decision can consider several plausible outcomes and the costs associated with each, using current, relevant information. That is a planning exercise, not a forecast or financial recommendation.
The most helpful salary figure is one with a clear population, definition, and date. Precision in the dollar amount cannot compensate for uncertainty about who the number represents or what effect the degree itself had.
