{"slug":"baseline-vs-endline-survey","title":"Baseline vs Endline Survey: Understanding the Difference","excerpt":"Baseline and endline surveys bookend a project. Together they reveal what changed. Here is how they differ, why both matter, and how to keep them comparable.","date":"2026-07-04","author":"DSTechnoverse","image":"https://images.unsplash.com/photo-1551288049-bebda4e38f71?auto=format&fit=crop&w=1200&q=80","category":"Baseline & Endline","tags":["baseline endline survey","endline","impact evaluation","M&E","monitoring"],"content":"In monitoring and evaluation, two surveys do the heavy lifting: the **baseline** and the **endline**. They sound similar and often use the same questionnaire, but they play different roles. Understanding both is essential to measuring impact.\n\n## The baseline: the starting line\n\nA **baseline survey** is conducted *before* an intervention begins. It captures the initial values of the indicators the project intends to change — the \"before\" snapshot.\n\n## The endline: the finish line\n\nAn **endline survey** is conducted *after* the intervention has run its course. It re-measures the same indicators to capture the \"after\" picture.\n\n## Why you need both\n\nThe magic is in the comparison:\n\n> Change = Endline value − Baseline value\n\n- The **baseline** alone tells you where people started.\n- The **endline** alone tells you where they ended.\n- **Together**, they reveal how much things changed over the project period.\n\nNeither is very useful without the other.\n\n## Keeping them comparable\n\nThe comparison only works if the two surveys are truly comparable. That means:\n\n- **Same indicators** measured the same way\n- **Same (or an equivalent) sample** from the same population\n- **Same questionnaire** wording, so responses mean the same thing\n- **Consistent methods** — mode, enumerator training, and timing considerations\n\nA change in wording, sampling, or method between baseline and endline can create \"differences\" that are artefacts of the survey, not real change.\n\n## Sometimes: a midline too\n\nLong programmes may add a **midline survey** partway through, to check progress and course-correct before the endline.\n\n## The causation caveat\n\nComparing baseline and endline shows *that* things changed — but not automatically *why*. Other factors (the economy, weather, other programmes) may also have moved the indicators. To attribute change to your project, you need a stronger design, such as a **comparison/control group** and methods like difference-in-differences.\n\n![Baseline, midline and endline compared](/blog-images/socio-economic/baseline-endline-comparison.svg)\n\n## The seven things that must be identical\n\nThe value of an endline lies entirely in its comparability with the baseline. Any drift between the two rounds shows up as change that the programme did not cause.\n\n| Element | What happens if it drifts |\n|---|---|\n| Question wording | A rephrased question collects a different answer, not a better one |\n| Response options and order | Adding or reordering categories redistributes responses across all of them |\n| Reference period | \"Last 30 days\" versus \"last month\" produce different recall |\n| Season of fieldwork | Agricultural income, migration and expenditure all move seasonally |\n| Household definition | Joint families and seasonal migrants must be classified the same way |\n| Mode of collection | Face-to-face and telephone elicit different answers on sensitive items |\n| Enumerator training content | Different probing conventions produce different distributions |\n\nThe mode row became a live issue for many studies whose baseline was face-to-face and whose endline was conducted by phone. Telephone respondents report differently on income, health and anything socially sensitive, and the shift is easily large enough to swamp a real programme effect. If the mode must change, run a subsample in both modes to estimate the difference.\n\nThe temptation to \"improve\" the questionnaire at endline is strong and should be resisted for any indicator you intend to compare. Add new questions if you wish; do not modify the old ones. A better question that breaks comparability is worse than a mediocre one that preserves it.\n\n## Attrition: the quiet destroyer of panel comparisons\n\nIf you track the same households across rounds, some will be gone by endline. The problem is not the loss itself — it is that the ones who leave are systematically different from the ones who stay.\n\n| Who tends to be lost | Direction of the resulting bias |\n|---|---|\n| Migrant and landless households | Endline sample looks wealthier than baseline |\n| Tenants and rented accommodation | Understates housing insecurity |\n| Households hit by a severe shock | Understates vulnerability |\n| Households that benefited most and moved on | Understates programme impact |\n| Households that never engaged | Overstates programme impact |\n\nThe last two pull in opposite directions, which is exactly why attrition cannot be waved away as random.\n\nThe standard diagnostic is simple and belongs in every endline report: take the baseline characteristics of households you successfully re-interviewed and compare them against the baseline characteristics of those you lost. If the two groups look alike, differential attrition is unlikely to be driving your result. If the lost households were poorer at baseline, say so and interpret accordingly.\n\nPrevention is cheaper than correction. Record GPS coordinates, head-of-household name, two mobile numbers, a neighbour's name and a landmark description at baseline. Tracking rates rise sharply with each of these, and none costs more than a minute of interview time.\n\n## Panel or fresh cross-section?\n\nThere are two legitimate ways to run an endline, and they answer different questions.\n\n| | Panel (same households) | Fresh cross-section |\n|---|---|---|\n| What it measures | Change within households | Change in the population |\n| Statistical power | Higher — each household is its own control | Lower for the same sample size |\n| Vulnerable to | Attrition bias | Population composition change |\n| Cost | Tracking effort | Fresh listing effort |\n| Best for | Programmes targeting specific households | Area-wide interventions |\n| Analysis | Household fixed effects, ANCOVA | Difference in group means |\n\nPanels are more powerful because each household serves as its own control, removing all the time-invariant differences between households from the comparison. That gain is often equivalent to a substantially larger sample. It is the right choice whenever the programme worked with identifiable households.\n\nA fresh cross-section is right when the intervention changed a place rather than a set of people — a village water system, a road, a market. There, in-migration and out-migration are part of the outcome, and holding the sample fixed would obscure it.\n\n## The midline decision\n\nA midline is worth its cost when the programme can still be changed. Its purpose is course correction, not impact measurement.\n\n| Add a midline when | Skip it when |\n|---|---|\n| Implementation runs three years or more | The programme is under 18 months |\n| Early results would change the design | The design is contractually fixed |\n| Attrition risk is high and needs monitoring | The population is stable |\n| Funders require interim reporting | No interim decisions are pending |\n| The delivery model is genuinely untested | The model is well established |\n\nA midline can be lighter than the full instrument — the key outcome indicators and the operational questions, not the whole demographic module. Halving the questionnaire halves the field cost and preserves the comparability of what matters.\n\n## Reading the comparison honestly\n\nThe gap between baseline and endline is not the programme's impact. It is the programme's impact plus everything else that happened in that period — a good monsoon, a new road, a national scheme, general economic growth.\n\nSeparating the two requires a comparison group, which is the subject of [how baseline and endline surveys power impact evaluation](/blog/baseline-endline-impact-evaluation) and, specifically, [difference-in-differences](/blog/difference-in-difference-baseline-endline). Where no comparison group exists, report the change as observed and say plainly that attribution cannot be established. That is a defensible position; claiming attribution from a before-and-after comparison is not.\n\nFor designing the first round see [what is a baseline survey](/blog/what-is-a-baseline-survey), and for keeping both rounds clean, [data quality control in large-scale surveys](/blog/survey-data-quality-control). Our [socio-economic and baseline survey service](/services/socio-economic-baseline-surveys) runs matched rounds with tracking built in. [3ie's evaluation resources](https://www.3ieimpact.org/) cover attrition analysis and panel design in depth.\n\n## The takeaway\n\nThink of baseline and endline as bookends. The baseline records where you began, the endline records where you finished, and the space between them is your story of change — provided you keep the two measurements consistent and interpret causation carefully.\n\n## Frequently Asked Questions\n\n### What is the difference between a baseline and an endline survey?\n\nA baseline measures conditions before an intervention begins; an endline measures the same indicators, in the same population, with the same instrument, after it concludes. The comparison shows what changed. Both are needed for any credible statement about change.\n\n### Should the endline survey the same households as the baseline?\n\nUse a panel of the same households when the programme targeted identifiable households — it gives more statistical power because each household acts as its own control. Use a fresh cross-section when the intervention changed a place, such as a road or a water system, and migration is part of the outcome.\n\n### What is attrition in baseline-endline studies?\n\nThe loss of households between rounds through migration, refusal or inability to trace. It matters because those lost usually differ systematically from those retained, which biases the estimated change. Always compare the baseline characteristics of retained and lost households and report the result.\n\n### Can I improve the questionnaire between baseline and endline?\n\nNot for any indicator you intend to compare. Changed wording, response options or reference periods break comparability and create apparent change that the programme did not cause. You may add new questions freely; do not modify existing ones.\n\n### Do I need a midline survey?\n\nOnly if the programme can still be changed in response to it. Midlines earn their cost on interventions running three years or more, where early findings can improve implementation. For short or contractually fixed programmes, they add cost without a decision to inform.\n\n### Does a difference between baseline and endline prove the programme worked?\n\nNo. The change includes everything else that happened during the period — weather, other schemes, economic trends. Attributing it to the programme requires a comparison group that experienced the same conditions without the intervention.\n\n\n---\n\n## How DSTechnoverse Runs Your Survey\n\nDSTechnoverse is a full-service socio-economic and baseline survey agency based in Indore. We do not just advise — we run the whole study end to end and hand you evidence you can defend:\n\n- **Design** — objectives, indicators, and sampling (random, stratified, cluster, or multi-stage), with a piloted questionnaire translated into local languages.\n- **Field data collection** — trained enumerators using CAPI (tablet) tools such as KoBoToolbox, ODK, and SurveyCTO-style apps, with GPS stamping, audio audits, and built-in skip-logic validation.\n- **Quality control** — back-checks, spot-checks, and daily high-frequency data monitoring so problems are caught in the field, not after.\n- **Analysis** — SPSS, R, Python, and Excel for indicators, cross-tabulations, and impact estimates, including baseline–endline difference-in-differences.\n- **A detailed report** — cleaned datasets, indicator tables, charts, GIS maps, and clear interpretation, with interactive Power BI / Tableau dashboards on request.\n\n**Our survey experience in Indore includes** Smart City citizen-feedback surveys, Startup India & MSME surveys, Swachh Survekshan (Swachh Bharat) cleanliness surveys, and skill-development impact evaluations — delivered for government, municipal, corporate, and NGO clients.\n\n**Planning a socio-economic, baseline, endline, or willingness-to-pay survey?** [Talk to our survey team](/contact) for a scope and quote.","related":[{"slug":"baseline-endline-impact-evaluation","title":"How Baseline and Endline Surveys Power Impact Evaluation"},{"slug":"what-is-a-baseline-survey","title":"What Is a Baseline Survey? A Complete Guide"},{"slug":"difference-in-difference-baseline-endline","title":"Measuring Impact: Difference-in-Differences with Baseline & Endline Data"},{"slug":"baseline-survey-for-csr-projects","title":"Baseline Surveys for CSR and Development Projects"}]}