Baseline surveys

Baseline surveys establish the social and economic conditions against which project impacts and subsequent changes can be assessed. For environmental and social risk management, they need to describe livelihoods, access to resources and services, social relationships, and the distribution of vulnerability across the affected population.

This section examines household survey design, fieldwork, and analysis through project questionnaires from Maluku, Indonesia, and census and survey evidence from Nepal. It brings those materials into comparison with national statistical instruments, with attention to household definitions, question wording, recall periods, sampling, and geographic coverage.

The baseline methods discussion addresses the assessment's purpose and evidence requirements. The articles below develop the survey itself: selecting respondents, defining variables, aligning questions with national standards, documenting field procedures, and producing estimates that support project decisions.

The dataset curation guide explains how to organize survey instruments, demographic sources, and geographic data for this work. The quantitative lab provides reproducible examples using synthetic household records.

Articles in this section

Establish what the baseline must explain

A description of average household conditions is rarely sufficient. Practitioners need to understand who may be affected, through which mechanisms, and with what consequences. A household may have a modest cash income but secure access to land, fishing grounds, reciprocal labor, and family support. Another may earn more but depend on a single employer or an unreliable rental arrangement. A survey that records only income would offer a limited account of either household's capacity to withstand change.

Social impact assessment connects material conditions with institutions, relationships, and people's own understandings of well-being. The IAIA guidance provides a useful methodological foundation for that broader inquiry (Vanclay et al., 2015; references). In practice, this means combining structured measurement with interviews, observation, seasonal calendars, and participatory mapping. Quantitative evidence establishes the distribution and magnitude of a condition; qualitative evidence helps explain how the condition matters.

Start with an impact pathway. For a road project, the pathway might be: land acquisition reduces productive land; reduced land changes crop production and labor demand; households adjust through wage work, borrowing, or migration. Each link calls for different information. Land area alone cannot establish livelihood loss, and livelihood loss alone cannot explain whether the proposed assistance is adequate.

Prepare an indicator specification before drafting questions:

These are proposed specifications, not universal indicators. Select measures that respond to the project's actual risks. Give each indicator a definition, question source, calculation, disaggregation, collection frequency, responsible party, and intended management response.

Distinguish the different surveys

Several activities are commonly called a baseline. They are related, but they serve different purposes.

IFC Performance Standard 5, paragraph 12, requires a census and appropriate socioeconomic baseline information where involuntary resettlement is unavoidable. Its 2023 handbook distinguishes the census, asset inventory, and socioeconomic survey in Module 4, Table 4.1. A sample survey of area residents therefore cannot stand in for identifying all displaced persons (IFC, 2012, 2023; references).

Record formal ownership, actual occupancy, tenancy, employment, customary access, and seasonal use separately. A land register may identify the registered owner while overlooking the tenant farmer, informal business, or person collecting resources from the same place. A household can appear in several impact records without becoming several households in the analysis.

Identify the applicable financing requirements in the survey terms of reference. ADB's Environmental and Social Framework became effective on 1 January 2026, while its Safeguard Policy Statement remains applicable to projects covered by the transition arrangements. Historical project documents and current requirements can therefore differ. Record the governing framework and version rather than treating every publicly available plan as an example of the same standard (ADB transition guidance).

Define the population in space and time

The project footprint is only one part of the survey boundary. Include places connected through access routes, water use, labor markets, construction facilities, host-community services, and resource dependencies where relevant. Distinguish direct exposure from other plausible pathways; do not label every nearby resident affected solely because of distance.

Specify the target population in ordinary language. For example: all households ordinarily resident in the defined study settlements on the listing date, plus a separately enumerated population of seasonal resource users. State how institutions, collective living arrangements, temporarily absent members, and people without a fixed dwelling will be covered. Those definitions determine both the sampling frame and the meaning of the estimates.

The reference date matters as much as the boundary. Collect the baseline before the relevant intervention wherever possible, and record earlier project activities that may already have changed behavior. Announcements, land speculation, compensation discussions, and construction mobilization can affect prices, expectations, or livelihoods before physical works begin. If the survey begins after these changes, describe it as a baseline for the remaining activities and document what cannot be reconstructed reliably.

Capture season explicitly. Interview date, agricultural stage, rainfall conditions, school term, and unusual disruptions can influence responses. A short survey conducted during harvest cannot simply be interpreted as an annual livelihood baseline. Use a seasonal calendar to select the survey window and identify measures requiring repeated observations.

Use public statistics to prepare better questions

Census tables, living-standards surveys, administrative records, and spatial data help identify plausible risks, benchmark definitions, and prepare fieldwork. They do not remove the need to observe conditions among affected people. A district poverty estimate is contextual evidence; it does not establish a particular household's poverty status or the adequacy of its compensation.

The Maluku discussion compares project questions with BPS household and village statistics. The Nepal discussion examines district profiles, census tables, and living-standards surveys. Comparisons require an explicit crosswalk between the project variable and the official measure: respondent, population, recall period, unit, response categories, geographic level, and collection date.

Maintain a source register, an indicator register, and a geographic crosswalk. Record the table, edition, and transformation supporting each result, with links to the issuing institution and a retained copy for subsequent review.

Retain differences that matter. A project may need to record access to a customary resource that a national household survey does not measure. Use a common core for comparable welfare measures and a local module for project-specific conditions. Harmonization should make relationships legible without removing the information needed to understand impacts.

Organize the work as an evidence chain

The survey design must connect questions, data, estimates, and decisions. The following sequence makes those connections reviewable:

1. Define the project decisions, impact pathways, target populations, and reference periods.
2. Compile existing evidence and identify its dates, geographic coverage, definitions, access conditions, and gaps.
3. Conduct preliminary consultation and qualitative inquiry to identify overlooked users, institutions, and local terminology.
4. Prepare the listing and sampling design, including procedures for absence, refusal, and frame errors.
5. Specify the questionnaire, data model, indicator calculations, and field manual together.
6. Test understanding, translation, interview burden, selection procedures, and digital logic; revise and freeze a documented version.
7. Collect data with supervision, daily quality review, and a record of interview outcomes.
8. Validate, estimate, triangulate, disclose appropriate findings, and assign actions and follow-up dates.

A pilot checks whether this chain works. It is not a small version of the final estimate. Include diverse livelihood, language, tenure, and access conditions; record where questions fail and why. Treat proposed interview-duration targets and back-check percentages as operational choices that need to be justified locally, not requirements imposed by an international standard.

Measure welfare and distribution together

No single variable represents the whole baseline. Combine a coherent welfare measure with measures of access, dependence, security, and social relations. For livelihoods, distinguish sales from profit, own-produced consumption from cash receipts, and recurring earnings from compensation or asset sales. For services, distinguish infrastructure presence from actual availability, affordability, and use.

Consumption can be useful for assessing material welfare, but it requires a defined aggregate and consistent valuation. The FAO and World Bank guidance addresses food-consumption measurement; the LSMS methods resources explain how recall periods, item lists, and respondent choice affect results (references). A brief spending question does not automatically reproduce an official poverty measure.

Report distributions alongside central values. A stable mean can coexist with substantial deterioration among tenants, households dependent on a restricted resource, or people requiring accessible transport. Pre-specify relevant disaggregations, retain sufficient observations for them, and explain when the data cannot support a reliable comparison. Present denominators and missingness beside percentages.

Connect the baseline to monitoring

Create persistent household and person identifiers before collection. Keep confidential contact details in a separate file and obtain specific permission for recontact. Decide how household splits, mergers, migration, death, and changes in membership will be recorded. Otherwise, apparent change may reflect a different population rather than different conditions.

A panel follows the same units. A repeated cross-section describes the population at each round. These answer different questions. For resettlement, continued attention to the original affected cohort is necessary even when people move away. For host-community monitoring, refreshing the sample may be needed to describe a changing population. Use both when the decisions require both.

Before-and-after differences describe change. They do not by themselves establish what the project caused. Inflation, rainfall, wider economic conditions, and other investments may explain some of the difference. The evaluation design needs a defensible counterfactual if causal attribution is the objective (Gertler et al., 2016; references).

The migration monitoring discussion applies these distinctions to settlement change, public-service pressure, and livelihood adjustment. Its spatial worked example also shows why a change in demand must be interpreted alongside access routes and local service capacity. Both are useful companions when a baseline extends beyond the project footprint.

Make findings usable and contestable

Return findings in formats people can understand. Explain what was measured, which groups were covered, and where uncertainty remains. Invite correction of factual errors and overlooked impacts without exposing individual responses. A public summary should be distinguishable from the confidential records used for case management.

The final baseline package should contain the questionnaire and translations; field and consent protocols; frame and selection documentation; data dictionary; indicator register; cleaned analytical tables; quality and nonresponse assessment; reproducible calculations; and a management table linking findings to responsibilities and deadlines. Keep raw records, corrections, and analytical transformations distinct.

The baseline then becomes part of project management: a documented account of conditions, a means of testing assumptions, and a basis for adjusting mitigation as circumstances change.

Applied analysis

The Upper Trishuli-1 cost analysis examines budgets, beneficiary denominators and evidence of delivery. The Bhutan demographic and spatial discussion connects census, survey and geographic data to exposure assessment. The social risk methodology connects these findings to financing decisions, supervision and verified outcomes.

Thematic Issues