The new economics of labour migration treats the decision to migrate as one taken by a household that spreads its members across places to spread its risks (Stark and Bloom, 1985). In the best-known statement of motives, Lucas and Stark (1985) distinguished altruism, self-interest and a middle position in which migrant and family are bound by an implicit contract of mutual insurance and repayment. Later reviews found that these motives are hard to separate in survey data, because the same transfer is compatible with several of them (Carling, 2008). Work in coastal Vietnam linked remittances to the resilience of households and also to widening inequality between them (Adger et al., 2002).

Anthropology starts from the gift. Mauss (1990) argued that a gift carries duties to give, to receive and to return, and Sahlins (1972) showed that the terms of reciprocity tighten as kinship distance grows, from open-ended giving among close kin to calculated exchange with strangers. Studies of relatedness add that kinship is made through feeding, housing and care as much as through descent, which means that sending money can itself create or confirm a tie. Zelizer (1994) demonstrated that people mark money for particular uses and relations, and Carling (2014) proposed that each remittance follows a recognisable script, such as allowance, gift, repayment or investment, that tells both parties what is owed. A design that records amounts without scripts misses what the money means to those who handle it.

Fishing economies have features that shape transfers. Catches are uncertain from day to day, earnings are divided by share and not by wage, crews are often recruited among kin, and credit from fish buyers ties households to particular landing sites (Acheson, 1981). Livelihood research has shown that fishing households nearly always combine fishing with other work, and that mobility is a normal part of that portfolio (Allison and Ellis, 2001). Erosion therefore acts on households that already move and already share. The research question is how an existing system of obligation is stretched by a new and permanent loss.

Fieldwork begins with residence in the village and with the ordinary work of participant observation: attending landings, net mending, auctions and household ceremonies (Bernard, 2011). Permission is sought from whatever body speaks for the village and then from each household individually. Early weeks are used to learn the local terms for kin, for kinds of help and for kinds of money, since the later instruments must use categories that people recognise. Questions about money are sensitive, and the ethical standards of the discipline require that people understand what will be recorded and can decline without consequence (American Anthropological Association, 2012).

A census records every household, its members present and absent, their work, the boats and gear they own or crew, and the condition and position of the house in relation to the shore. Genealogies are then collected by the method that Rivers (1910) set out more than a century ago, extended to include affines and absent members and their present whereabouts. The product is a kin chart for each household that links to the charts of others. This chart is the frame on which transfers will later be placed, and it identifies migrants to be traced in the fourth phase.

One-off questions about remittances invite rounded and selective answers. The design instead uses diaries, adapted from the financial diary method developed for studying the money management of poor households (Collins et al., 2009). A purposive selection of households, chosen to span those that have lost a house, those at risk and those further inland, is visited every two weeks. At each visit the researcher records every transfer received or sent since the last one: who sent it, the relation, the channel, the amount in broad bands, what was said about its purpose and how it was used. Requests made and refused are recorded as carefully as money that arrived.

Towards the end of the diary period every household in the census is asked a short set of name-based questions: from whom it received help in the past year, to whom it gave, and whom it could ask in an emergency. The answers are assembled into a network in which households and absent migrants are points and transfers are directed ties, following standard practice in network research (Wasserman and Faust, 1994). Laid over the genealogies, the map shows which categories of kin carry support and where non-kin, such as boat owners and fish traders, stand in.

Remittance is a relationship with two parties, and the village holds only one of them. Following the argument for multi-sited ethnography (Marcus, 1995), the researcher traces a number of migrants from diary households to the towns, harbours or foreign worksites where they live, and interviews them there or, where travel is impossible, by telephone. Sender and receiver are asked about the same transfers: how the request was made, what was felt to be owed, what was sent, and what each thinks the other did with it. Differences between the two accounts are the main material for the second research question. Life-history interviews with older residents record how obligations were met before migration became common.

Residents are asked to walk the shore with the researcher and to mark on a sketch map where houses, wells, shrines, net-drying ground and landing places stood, and when each was lost. The map dates erosion from the village's own memory and ties each loss to a household in the census. Analysis then brings the strands together for each diary household: its position on the kin chart, its transfers through the year, the scripts attached to them, and the use of the money in relation to its losses. Findings are presented back to the village in an accessible form before anything is published, and corrections are invited.