Household Heads and a Cash Transfer
An NGO is designing a three-year conditional cash transfer for school attendance in a region where the dominant ethnic group is matrilineal: lineage membership, inheritance of groves, and house rights pass from mother to daughter, while day-to-day authority over children is exercised by the mother's brother. The NGO's baseline survey used a standard household roster: 412 households, each with a 'male household head' bank account. Enumerators reported that in 38 percent of compounds the person who actually decides school spending is a woman - a mother, or a maternal uncle's sister - who is not the account holder. The government's social registry, which will receive the data, uses household headship to assign land-tax categories. The implementing partner wants a clean dataset with no missing head-of-household fields, and the donor wants an impact estimate with a household-level treatment indicator.
1. What is wrong with the household measure, and how should the unit be redefined?
Read the worked answer
The roster measures a legal-administrative unit, not the unit that makes schooling decisions. Three failures:
- Misplaced headship: where avunculate authority operates, the mother's brother may control cash for fees while the mother controls daily provisioning. A 'male household head' field captures one and erases the other; in 38 percent of compounds the actual decision-maker is a woman who is not the account holder.
- Residence versus membership: compounds are not nuclear households. Children may be fostered to a grandmother or an uncle, and lineage members may sleep elsewhere. A roster keyed to co-residence misassigns them.
- Unit of treatment: a transfer paid into one account treats a compound as a single decision unit, which is empirically false.
Fix: map decision rights before finalizing the instrument - who decides enrollment, who pays fees, who can withdraw a child - then define the recipient unit explicitly and record multiple decision-makers instead of forcing one head.
2. How should the payment be delivered so that it reaches the person who decides school spending?
Read the worked answer
Payment design should follow the decision-rights map, not the registry.
Options:
- Pay the mother or the enrolled child's primary caregiver directly by mobile money, with an account-opening clinic; in 38 percent of compounds this reaches the actual spender.
- Where a maternal uncle controls fees, allow a co-signatory or a split payment (for example 70 percent to the mother, 30 percent to a fee account) rather than routing everything to the male head.
- Condition on enrollment and attendance verified at school, not on the identity of a 'head.'
Risks:
- Backlash: directing cash to women can provoke intra-household conflict; consult women's groups and monitor.
- Elite capture: mobile accounts may be captured by phone-owning men; use group verification or a witness at enrollment.
- Exclusion: fostered children and compounds with absent migrants may drop out of the frame; add a proxy-respondent rule.
Pilot the modality in two villages and measure receipt by the intended decision-maker before scaling.
3. What conditions should govern sharing the dataset with the government registry?
Read the worked answer
This is a secondary use with serious consequences: headship data feed land-tax categories, and in a matrilineal area a misrecorded head can affect a lineage's claim to groves and house rights. That is not a routine data transfer.
Conditions:
- Purpose limitation: share only what the program requires; no full rosters with names and land identifiers.
- Consent that names the government use, in plain language and the local language. Consent to program enrollment is not consent to tax assignment.
- Aggregate or pseudonymize before transfer, with the linkage key held by the NGO or an independent custodian.
- A correction process so households can contest a headship or land record.
- A risk assessment: if the data could trigger tax reassessment or land disputes, delay the transfer until protections exist, and tell the donor why.
The donor's demand for a clean household-level treatment indicator does not justify supplying a dataset that could be used against the people who generated it.