The entitlement matrix: the document that decides everything else

1. Why the matrix governs

The matrix answers, in one table, the question every subsequent process depends on: who gets what. The census instrument must collect the data needed to place each household in a category. The budget is the matrix priced against the census. The payment record is an assertion that a matrix entitlement was delivered. A grievance is, almost always, a claim that the matrix was applied incorrectly. [3] The audit tests entitlement against delivery.

Because everything derives from it, defects propagate. A category omitted from the matrix is a category the census does not collect data on, which means it cannot be added later without re-surveying, and re-surveying after a cut-off raises its own difficulties. The matrix is therefore among the few documents on a project where an error made in a week of drafting cannot be corrected by any amount of subsequent effort.

This is worth stating plainly because matrices are frequently produced by adapting one from a previous project — a practice the World Bank's own project standards presuppose will happen and try to discipline through the ESCP rather than prevent. [5] Adaptation is sensible — the structure is stable across projects — but the categories must be tested against the tenure and livelihood realities of this location, and it is precisely that testing which adaptation tends to skip.

2. Categories of person, not categories of asset

A matrix has two axes: the loss and the person. The loss axis is comparatively easy — land, structures, crops, trees, business income, access. The person axis is where matrices succeed or fail, because it determines who is recognised at all.

The standard's position is that entitlement to compensation for land depends on rights in land, while entitlement to resettlement assistance does not — PS-5's third eligibility category exists for exactly this population. [1] Guidance Note 5 is explicit that the absence of a recognisable right does not remove the entitlement to assistance. [2] People occupying without recognisable legal right are not entitled to be paid for the land, and are entitled to assistance for what they lose and to be no worse off. [1] The World Bank's ESS-5 draws the same distinction between compensation and assistance by tenure category. [4] A matrix that has no row for them has silently decided otherwise.

The categories most often missing are consistent across projects: tenants and sharecroppers, whose loss is a tenancy rather than a title; agricultural labourers, whose loss is employment with no asset at all; users of common resources; adult children within a household who farm separate plots but are enumerated as one unit; and business operators without premises. Each is invisible to a census designed around the titled landowner.

3. The household as unit

Matrices operate on households, and household is an administrative construct rather than an observed fact. Where the census records one head per household and compensation is paid to that head, the matrix has made a distributive decision affecting everyone else in the unit without stating it as a decision.

The consequences are documented and predictable. Women's interests in land — often use rights rather than ownership, often unregistered — are extinguished by a payment to a male household head. Polygamous households, extended households, and households with absent members are compressed into a single record that does not reflect how the asset was actually used or how the compensation will be shared.

Matrices that address this do so explicitly: joint entitlement or joint payment requirements, separate recognition of use rights, definitions of household that permit splitting where sub-units farm separately. These provisions are administratively burdensome and they are the difference between compensating a household and compensating whoever the enumerator wrote down.

4. Design failures that cannot be undone

Three failures recur and share the property of being uncorrectable after implementation begins. The first is a missing category, discussed above: no data was collected, so no retrospective determination is possible.

The second is an ambiguous entitlement — an entry providing 'assistance to restore livelihoods' or 'support to relocate' without specifying what is provided. At the point of implementation this is resolved by whoever is administering payment, differently for different households, and the resulting inconsistency is both a grievance generator and indefensible at audit.

The third is a matrix inconsistent with the budget. Where the matrix promises replacement housing and the budget provides cash at a rate below construction cost, the inconsistency will be resolved in favour of the budget, and the matrix's promise becomes the basis of a grievance that is entirely well-founded. Pricing the matrix against the census before approval is the check that catches this, and it is frequently done after approval or not at all.

5. Conclusion

IFC's Guidance Note 5 sets out worked examples of matrix categories precisely because the categories are where projects most often fail. [6] The entitlement matrix is a short document whose drafting decisions are effectively irreversible, and it is routinely produced under time pressure by adapting a template. The amounts in it attract scrutiny; the category structure, which determines who is eligible for any amount at all, usually does not.

The most useful discipline is to read the matrix against a list of the people actually present in the affected area — not the titled owners, but everyone whose living depends on the land — and to ask which row each of them falls into. The people who fall into none are the people the programme will fail, and they can be identified before anything has been paid.