What a baseline study actually tells you — and what it cannot

Baseline studies occupy an odd position in the programme cycle. Everyone agrees they are necessary — funders require them, appraisal committees ask for them, evaluation frameworks assume them — and yet they are routinely commissioned late, read once, and filed. Part of the problem is that the people commissioning baselines and the people reading them often expect the document to do things it was never designed to do.

This post sets out, plainly, what a well-conducted baseline can establish, what it structurally cannot, and how to commission one so that it earns its cost three years later, when the endline team arrives and needs something solid to compare against.

What a baseline is for

A baseline answers one question: what is the situation on the indicators this programme intends to move, before the programme starts moving them? That is the whole job. Everything else — sampling strategy, instrument design, enumerator training, data quality protocols — exists in service of that question.

Done properly, a baseline gives you three things.

First, a defensible starting measurement. If your programme claims it will improve institutional delivery rates, or reduce dropout among adolescent girls, or increase uptake of a social protection entitlement, the baseline is the number your endline will be judged against. If the starting measurement is soft, every claim you make later inherits that softness.

Second, a reality check on programme design assumptions. Programme documents are written in offices, often months before anyone visits the field. A baseline frequently reveals that the assumed problem is real but differently shaped — the entitlement is known but the paperwork is the barrier; the ASHA worker is trusted but overloaded; the gram panchayat is willing but has never been asked. Design corrections made at this stage cost almost nothing. The same corrections made after a midline cost a great deal.

Third, a sampling and measurement architecture for the whole results chain. The baseline is where you decide, concretely, how each indicator will be measured, from whom, and at what level of disaggregation. Those decisions bind the midline and endline. A baseline that measured "awareness" with one loosely worded question has quietly decided that the endline can never say anything precise about awareness.

What a good baseline tells you

Read carefully, a competent baseline report should let you answer questions like these:

• What is the current level of each outcome indicator, with a stated margin of uncertainty?
• How do those levels differ across the groups the programme cares about — by block, by social category, by gender, by landholding, by whatever axis the intervention will work along?
• Which assumed barriers actually show up in the data, and which do not?
• What does the service delivery environment look like — staffing, supply, distance, functionality — independent of household-level demand?
• What should the programme realistically expect to shift, given where the numbers start?

That last point deserves emphasis. A baseline that finds an indicator already at a high level is telling you something important: the headroom for improvement is small, and the endline comparison will be noisy. A programme team that reads its baseline honestly will sometimes revise its targets. That is not weakness; that is the baseline doing its job.

What a baseline cannot tell you

Here the misunderstandings begin, so it is worth being blunt.

A baseline cannot tell you whether your programme will work. It describes conditions; it does not predict responses to an intervention that has not yet happened. Treating high need as evidence of likely impact is a category error — the districts with the worst indicators are often the ones where delivery is hardest.

A baseline is not an evaluation of anyone. District administrations and implementing partners sometimes read baseline findings as an audit of their performance and respond defensively. A baseline measures a situation produced by decades of geography, economics, and administrative history. Commissioners should say this explicitly at dissemination, because a defensive line department is a poor programme partner.

A baseline cannot establish causes. It can document that open defecation persists where toilets exist, or that girls drop out at a particular transition point. It can gather stated reasons. But stated reasons are hypotheses, not findings. Causal claims need evaluation designs, not description.

A baseline cannot substitute for monitoring. It is a photograph, not a film. Programmes that skip routine monitoring because "we have a baseline and will have an endline" discover at endline that they cannot explain their own results.

Where baselines go wrong in practice

Three failure modes account for most wasted baseline budgets.

1. Commissioned after implementation has begun. Once field activity starts, the "before" picture is contaminated, and no amount of statistical apology fully repairs it. If the timeline has slipped, say so in the report rather than pretending otherwise.
2. Instruments copied from other studies without adaptation. A module that worked in one state may misfire in another — terms for entitlements differ, seasonal calendars differ, what counts as a sensitive question differs. Piloting in the actual study area is not a luxury.
3. No custody plan for the data. Staff change, agencies change, and five years later nobody can find the cleaned dataset or the codebook. Insist on receiving the raw data, the cleaned data, the instruments, and the sampling documentation — not just the report.

Commissioning a baseline that will still matter at endline

If you are drafting terms of reference, three requirements repay their weight many times over. Fix the indicator list and its definitions before fieldwork, in writing, with the endline explicitly in mind. Require full documentation of sampling and non-response, because the endline team will need to replicate both. And build in a structured discussion of findings with the programme team — not a dissemination event, a working session — so the baseline changes decisions rather than shelf space.

A baseline is a measurement instrument for an argument you will need to make years from now. Commission it with that argument in mind, and it becomes one of the cheapest forms of insurance a programme can buy. Our research practice is built around exactly this discipline.

If you are planning a baseline, or trying to make sense of one you have inherited, we would be glad to talk it through — write to us.