Measurement Choices Are Policy Choices

Indicators look like description and behave like allocation. Once a system reports on foundational learning, or on school resourcing, or on learning inequality, budgets and attention start to move toward what the indicator can see. Everything it cannot see becomes, in practice, less real.

I have worked on both sides of this. Building comparable foundational learning estimates across many countries means making decisions about which surveys count, how to weight them, and what threshold separates a child who can read from one who cannot. Each decision is defensible. Each also changes which countries appear to be improving.

The same is true of inequality measures. A framework that summarises learning inequality across contexts, like the Gini Learning Index work I contributed to, is valuable because it makes very different systems commensurable. It is also a claim about what kind of inequality matters most, and it will be read as such by ministries deciding where to intervene.

This is not a reason for researchers to retreat into caveats. Systems will measure something whether or not methodologists help. It is a reason to design indicators with the political economy in view: to publish the sensitivity of a ranking to its assumptions, to keep the underlying dimensions visible rather than only the headline number, and to be explicit about what the indicator was never built to capture.

The most useful research I have been part of has been the kind that gives a decision maker both a number and a clear sense of its edges. Rigour is not only about precision. It is about handing over a measure that the next person can use responsibly.