Questionnaire design and fieldwork
A baseline questionnaire should produce evidence that can be interpreted, compared, and collected again. Define each measure alongside its questions, calculation, respondent, and field procedure. The survey design method establishes the assessment purpose; this guide develops the interview and data specifications.
For national comparisons, use the relevant official questionnaire and methodological notes. The Indonesia guide connects the Maluku project question sets with BPS household and village statistics; the Nepal guide covers census and living-standards evidence. Match the household definition, eligible respondent, recall period, units, codes, and calculation. Add questions on local resource use, rights, exposure, and project expectations where the assessment requires them.
Specify the observation and respondent
Distinguish households, people, dwellings, parcels, assets, enterprises, and resource-use activities. Several households may occupy one dwelling or share a productive asset. Linked records preserve these relationships and support consistent denominators.
| Table | Record identifier | Main relationships |
|---|---|---|
| Household interviews | Household ID and round | Listing, interview outcome, respondent, survey version |
| Members | Household ID, person ID, and round | Membership, age, schooling, work, functioning |
| Dwellings | Structure ID and observation date | Occupants, construction, services, location |
| Parcels and assets | Parcel or asset ID and date | Location, use, measurements, affected extent |
| Rights and uses | Relation ID | Person or household linked to asset, with type of claim or use |
| Livelihood activities | Activity ID, household ID, and period | Production, sales, costs, labor, resource access |
| Consumption items | Household ID, item ID, and period | Acquisition source, quantity, unit, value |
Keep persistent analytical identifiers and date changes in membership, residence, or use. Store names, national identity numbers, and precise locations separately where confidentiality requires it.
Choose respondents according to the information sought. A knowledgeable household member can describe shared expenditure and housing; questions about personal working conditions, control over assets, or participation in decisions often require an individual interview. Record proxy responses so that their source remains identifiable.
Develop a common core and local modules
A common core supports repeat measurement and specified comparisons. Local questions address particular impact pathways: landing places and seasonal access for a coastal project, or parcel use, crop cycles, and easement restrictions for a transmission line.
For every question, specify its variable ID, analytical purpose, wording, respondent, eligible population, reference period, response type, allowable codes, unit, skip rule, validation rule, calculation, translation, and source. Stable identifiers such as water_roundtrip_minutes allow questions to move within the form while retaining their analytical meaning.
Separate current conditions, expectations, preferences, and agreements. Record the project information available to respondents when asking about anticipated employment, disruption, or relocation. Formal consent, eligibility, and entitlement each require their own applicable procedure and evidence.
Make quantities comparable
Record the quantity, period, and unit for every numerical response. Retain the reported local unit alongside any standardized value, conversion factor, and supporting measurement. Commodity and geographic variation matter when converting baskets, bundles, land units, or catch quantities (Oseni et al., 2017; references).
Specify the accounting method for income. Record wages and pay periods; enterprise receipts and operating costs; agricultural production, sales, and costs; remittances; and other transfers separately. Calculate net earnings from the relevant receipts and costs, and distinguish cash flows from own-consumed production. Allocate shared household and business expenses consistently.
Consumption questions should cover purchased, own-produced, and received food, including meals away from home. Define valuation rules and item-specific periods for nonfood expenditure, particularly education, health, and durables. The FAO and World Bank guidance examines these choices and their effects on measurement (references).
One interview design uses seven days for food, a defined recent month for routine nonfood costs, and twelve months for selected irregular expenditure. Select and test periods against the intended official comparison and local seasonality. Annual estimates require an explicit treatment of seasonal variation and irregular purchases.
Represent absence and uncertainty explicitly
A genuine zero is an observed value. Record the reason for a missing value separately so that nonresponse and eligibility remain visible.
| Status | Meaning | Treatment |
|---|---|---|
observed |
A reported or measured value, including zero | Include in the defined calculation |
not_applicable |
Outside the eligible question population | Exclude from its eligible denominator |
dont_know |
Respondent could not provide the value | Report item nonresponse |
refused |
Respondent declined | Report separately |
not_asked |
Omitted by design or error | Check the form and skip path |
pending_verification |
Provisional observation awaiting review | Resolve or disclose its status |
Store missing numerical values as nulls, with the status in its own field. For example, confirmed absence of agricultural activity can justify zero agricultural earnings; refusal to report earnings remains missing.
Test language and interpretation
Review translations with speakers who understand the survey concepts and local usage. Use cognitive interviews to ask how participants interpreted a question, which period they considered, and how they arrived at an answer. Independent review and back-translation help identify wording differences; respondent testing establishes how questions work in practice.
Test household membership, ownership, occupation, dependence, and community. Relevant probes include support from relatives living elsewhere, borrowed land, and seasonal fishing. Retain open-text explanations and the respondent's language where these clarify categories or meanings. Use the pilot to refine wording, response codes, and instructions.
Collect sensitive information proportionately
Explain the survey's purpose, commissioning organization, voluntary participation, intended uses, and confidentiality arrangements. Record consent to the interview, recontact, photographs, audio, and geolocation separately, and allow respondents to decline individual questions. Explain the distinction between participating in research and agreeing to a project transaction.
For comparable disability-related functioning measures, use the Washington Group Short Set's six domains, wording, and response categories. Address additional support needs through separately specified questions (references).
Collect identifiable violence histories only through a dedicated ethical and safety protocol with trained staff, privacy, secure records, and referral arrangements. WHO's Putting Women First provides guidance for designing that research safely (references).
Implement a documented field protocol
Train enumerators through practice interviews covering neutral probing, consent, units, respondent selection, local conversions, interview outcomes, and device security. Discuss how project sponsorship and interviewer identity may affect responses. Arrange interviews so participants can speak freely.
For every selected unit, record completion, partial completion, refusal, temporary absence, ineligibility, duplicate listing, or inability to locate. Schedule callbacks and follow a pre-specified statistical procedure for any reserve sample. Preserve the disposition of each selected household.
Digital forms support conditional questions, repeats, range checks, and calculations. ODK's form-logic documentation describes their implementation. Apply hard constraints to impossible entries and warnings to unusual but plausible values, with explanations available for review.
Review interview duration, item nonresponse, repeated answers, coordinates, roster consistency, duplicate IDs, and patterns by enumerator each day. Investigate anomalies using field context and consented back-checks of stable information. IFC's baseline-data guidance also covers piloting, interview consistency, and testing data systems.
Preserve a correction trail
Keep raw submissions unchanged. Log corrections with the record ID, variable, previous and revised values, reason, evidence, reviewer, and date. Generate the cleaned dataset through documented transformations, retaining the basis for exclusions and outlier treatment.
Before analysis, reconcile roster counts, interview outcomes, sample selection, geographic coverage, units, and indicator denominators. Separate confidential case records from analytical and public datasets. Report remaining uncertainty and missingness with the estimates they affect.