Baseline

“A baseline is not the past in general; it is the particular past against which a claim chooses to be judged. ”

Baseline sounds like a neutral starting point. In practice, the choice of baseline can determine whether progress appears ambitious, modest or nonexistent. Select an unusually bad year and almost any later performance improves. Select an unusually good year and genuine progress can disappear.

The OECD evaluation glossary describes baseline data as information collected before or at the start of an intervention against which progress can be assessed. In greenhouse gas accounting, a base year performs a related function for inventories. Project accounting may instead use a baseline scenario representing what would probably have occurred without the intervention. These uses should not be collapsed.

A historical baseline records an observed condition at a defined time. A counterfactual baseline estimates an alternative condition that was not observed.

A performance benchmark sets an external reference. Each supports a different claim and carries different uncertainty. Consider farmer income. A programme may compare household income after training with income in the drought year immediately before implementation. The increase is real relative to that year, but part may reflect rainfall returning to normal.

A multi-year baseline or comparison group may provide a fairer reference. The baseline does not remove the need for causal reasoning. Climate change makes some baselines unstable. Historical rainfall, pest patterns and yield distributions may no longer represent the conditions a programme will face. Infrastructure designed against a twentieth-century flood baseline can fail in a changing climate.

Adaptation planning may require forward-looking reference scenarios rather than a fixed historical average. Boundaries matter equally.

A company can reduce emissions per tonne while total production grows and absolute emissions rise. An intensity baseline and an absolute baseline answer different questions. A forest programme can improve tree cover inside project boundaries while displacement occurs outside. The reference must match the system and claim. Data quality at the baseline cannot be assumed.

Early measurements are often weaker because systems were not built yet. Later monitoring may use better instruments, revised definitions or more complete coverage. Comparing the two without harmonisation can turn a measurement change into apparent performance. Recalculation rules are therefore important.

Organisational changes, acquisitions, methodology revisions or discovery of material errors may justify restating a baseline. Silent changes undermine trust. A credible system states when recalculation occurs and preserves the original and revised values with reasons. Baseline selection can also have distributive consequences. National averages may conceal groups starting far behind.

A programme can meet an average improvement target while the lowest-income households deteriorate. Disaggregated baselines show different starting conditions and prevent equality of treatment from being mistaken for equity of outcome. A baseline is not automatically a target. It tells us where the reference begins, not where performance should end. Nor is it a prediction.

The counterfactual or scenario may evolve as external conditions change, but changes should follow pre-defined rules rather than the result desired. Different decisions may require different baselines.

A climate target may use a fixed historical year so progress can be tracked consistently. A project evaluation may need a contemporaneous comparison because prices, rainfall and policy are changing. A restoration programme may need both an ecological reference condition and a social baseline describing access, use and livelihoods.

Treating one reference as sufficient for every purpose can make the numbers internally consistent but analytically weak. Baseline governance should therefore address revision. Data errors may be discovered, boundaries may change and improved methods may become available.

A revision can be legitimate, but it should be documented, applied consistently and accompanied by a restatement where comparability would otherwise be lost. A baseline that changes only when performance looks poor is not improved evidence; it is moving the starting line.

The discipline is to treat the baseline as part of the claim. State the period, population, boundary, indicator, method, uncertainty and reason for selection. Ask whether another defensible baseline would materially change the conclusion. Progress is credible only when the reference is credible too.

Practical application

Choose baseline periods that reflect normal variability and the intended decision. Document boundaries, definitions, methods, coverage and external conditions. Use multi-year or modelled references where single-year values are unstable, and disaggregate groups with different starting conditions.

Establish recalculation rules before results are known. Preserve version history and explain changes. Distinguish historical baselines, counterfactual scenarios and benchmarks in reporting. Maintain a baseline register with the approved value or scenario, source data, calculation, boundary, owner, approval date and revision rule.

When a baseline changes, retain the earlier version and show the effect of restatement on every linked target and trend. This prevents one correction in a spreadsheet from silently changing years of reported progress.

Why it matters

Every claim of improvement depends on a reference. Weak or opportunistic baselines can manufacture progress, hide deterioration and make targets incomparable across programmes or organisations.

Common misconception

A baseline is often treated as whatever value was available before a programme began. It should be a fit-for-purpose reference whose period, boundary and method support the claim being made.

Connections

Indicator defines what is measured, Target defines the desired future value and Benchmark provides an external comparison. Counterfactual estimates what might have happened without the intervention, while Attribution tests the relationship between intervention and change.

A question worth asking

Would your performance claim still look convincing if the baseline were chosen before anyone knew the result?

Selected references

OECD. 2023. Glossary of Key Terms in Evaluation and Results-Based Management for Sustainable Development, Second Edition. Kusek, J. Z. and Rist, R. C. 2004. Ten Steps to a Results-Based Monitoring and Evaluation System. Greenhouse Gas Protocol. 2004. The GHG Protocol for Project Accounting. Greenhouse Gas Protocol. 2004. A Corporate Accounting and Reporting Standard, Revised Edition.

Lamhauge, N. , Lanzi, E. and Agrawala, S. 2012. Monitoring and Evaluation for Adaptation: Lessons from Development Co-operation Agencies. OECD.