Monitoring a resettlement: indicators that measure outcomes, not activity

Most resettlement monitoring reports count activity. Households compensated, payments disbursed, meetings held, training sessions delivered. These are easy to collect, arrive early, and rise reassuringly over time.

A completion audit asks a different question: are these people no worse off than they were?[1] Activity counts cannot answer it, and a monitoring system designed around them will run for the whole project and produce nothing the audit can use.

Three layers, and only one of them is optional

Useful resettlement monitoring separates three things that are routinely collapsed into one report:

Activity monitoring is the layer that can be dropped without loss. It is also the layer most projects report almost exclusively.

Compliance indicators worth tracking

These are the ones a completion auditor will reconstruct if you have not tracked them — and reconstruction is where projects discover problems too late to fix. Each is a percentage with a defined denominator:

• Compensation before displacement. Share of households whose payment date precedes their parcel's displacement date. Requires both dates recorded per household, which is the part usually missing.
• Payment completeness. Entitlements determined versus entitlements discharged, reconciled against disbursement records rather than against the payment instruction.
• Register growth. People added after the census, with reasons. A flat line here usually means the grievance mechanism is not reaching anyone.
• Grievance timeliness. Distribution of days to resolution, and the count breaching the service standard — not the total received.
• Vulnerability screening coverage. Share of households screened against written criteria, including recorded negatives.
• Disclosure coverage. Share of affected communities with a documented disclosure event, by section.

Outcome indicators, and the baseline problem

Outcome indicators are only meaningful against a baseline, which is why the socioeconomic survey has to be designed backwards from them.[2] If income is going to be tracked by source at 12 and 24 months, income by source has to be captured at census — in the same categories, by the same definitions.

Projects that design monitoring after implementation begins discover their baseline recorded something adjacent to what they now need to measure, and the comparison cannot be made. There is no recovery from this. The baseline moment has passed.

Every outcome indicator is a promise made to the census. If the census did not collect it, the indicator does not exist.

Practical outcome indicators for a resettlement, disaggregated by livelihood segment and by sex:

• household income by source, versus baseline;
• productive asset holdings — land under cultivation, livestock, business stock;
• yield per unit area for land-based households;
• businesses still trading at 12 and 24 months post-relocation;
• housing quality and security of tenure at the replacement site;
• access to water, school and health facilities versus baseline;
• food security through the lean season.

Sampling, panels and attrition

Compliance indicators need full coverage — a percentage of households paid before displacement is meaningless as a sample. Outcome indicators can be sampled, and normally should be, because the depth of a livelihood survey makes census-wide repetition impractical.

Two design points decide whether that sample is usable. Use a panel — the same households at each round, so change is measured rather than inferred from different populations. And track attrition, because the households that drop out are not random. Those who moved away, gave up, or migrated are disproportionately the ones whose restoration failed, and a panel that quietly loses them reports rising averages while the outcome deteriorates.

Who monitors

Internal monitoring by the implementing agency runs continuously and feeds management decisions. External monitoring, by a party independent of delivery, provides the verification a lender relies on — and the completion audit is a distinct exercise again, conducted after implementation is substantially complete.[3]

The three are frequently conflated in RAP budgets, with a single line for "monitoring" that funds none of them adequately. They have different purposes, different independence requirements and different costs, and they should be budgeted separately.

Monitoring that changes something

The final test of a monitoring system is whether any finding has ever caused the project to do something differently. A system producing reports nobody acts on is an expense, not a control.

In practice that means defining, in advance, what triggers a response: an income indicator below baseline at 12 months triggering a review of livelihood measures; a grievance cluster triggering re-examination of the rate that produced it; transitional support extending because recovery indicators say it must.[4] Triggers written down before the data arrives are what separate monitoring from reporting.