Replacement Procedures

Replacement procedures differ fundamentally between baseline and follow-up rounds. This guide explains when replacement is permitted during baseline data collection, the critical protocols to minimize selection bias, and why replacements are never allowed in any follow-up round of panel studies.

TipKey Takeaways
  • Baseline replacements are permitted only for deceased, permanently moved, ineligible, or inaccessible respondents—always using pre-generated random replacement lists, never enumerator-selected.
  • Follow-up replacements are strictly prohibited in any panel study to preserve the ability to measure individual-level change over time.
  • Intensive tracking strategies and mop-up operations are the only acceptable approaches to minimize attrition in follow-up rounds.

Replacement procedures differ fundamentally between baseline and follow-up rounds due to the requirements of panel data structure.

Baseline Replacements

During baseline data collection, researchers may replace ineligible or unreachable respondents following strict protocols to maintain sample validity and minimize selection bias.

When can researchers replace respondents?

Replacement is only permitted under specific circumstances where the original respondent cannot participate:

  • Respondent deceased before first contact: Individual died before baseline survey could be conducted
  • Respondent permanently moved away: Confirmed untraceable after exhaustive tracking efforts
  • Respondent ineligible: Does not meet study inclusion criteria upon verification
  • Location inaccessible: Security or safety concerns prevent enumerator access to the area

Critical Protocols

These five principles ensure baseline replacements maintain rigor and minimize selection bias.

  1. Random allocation required: Replacements must come from randomly-generated replacement list, never selected by enumerators
  2. Same characteristics: Replacement should match original respondent’s key stratification variables (e.g. region, treatment status, etc.)
  3. Document thoroughly: Record why original respondent replaced and ID of replacement
  4. Supervisor approval: All replacements require field supervisor authorization
  5. Maintain independence: Replacement respondents must be independent selections, not neighbors or referrals
WarningWarning: Convenience Sampling Risks

Never allow enumerators to select their own replacements. This introduces severe selection bias as enumerators naturally choose:

  • Nearby respondents that are convenient to reach, creating geographic bias
  • Friendly, cooperative individuals, creating cooperation bias
  • Respondents who validate enumerator expectations, creating confirmation bias

Always use pre-generated random replacement lists with specific IDs assigned to each potential replacement scenario.

Follow-up Round Replacements

Panel studies have fundamentally different replacement rules than baseline surveys. Understanding and following this rule is essential for maintaining valid longitudinal data.

WarningCritical Rule for Panel Studies

Never replace respondents in any follow-up round

This rule applies to:

  • 2-round studies—baseline + endline
  • 3-round studies—baseline + midline + endline
  • Any study with two or more rounds—including long-term panels

Why this matters?

Replacing respondents in follow-up rounds destroys the panel structure that is the entire purpose of longitudinal studies. Panel data allows researchers to:

  • Measure individual-level changes over time
  • Control for unobserved individual characteristics
  • Identify causal effects with greater precision

A panel dataset with high attrition is problematic, but a panel dataset with replacements is worthless because it can no longer measure true individual change.

TipInstead of Replacement: Intensive Tracking

When respondents are difficult to locate in follow-up rounds:

  1. Exhaust all tracking strategies before considering cases as closed. See Longitudinal Tracking Strategies.
  2. Budget for tracking efforts: Allocate 15-25% of field budget specifically for difficult cases
  3. Use specialized mop-up teams: Deploy experienced trackers for hard-to-find respondents. See Mop-Up Operations section.
  4. Document attrition carefully: Record reasons for non-response to assess potential bias
  5. Analyze differential attrition: Check whether attrition differs between treatment groups
Back to top