This reflexion focuses on the connection between anthropological fieldwork, quantification and theory - arguing that anthropologists should not content themselves with locally valid explanations, but should contribute to general theories. The paper originated as part of the "round table discussion" at the Workshop on the Anthropological Demography of Europe, held in September 2005.
Fieldwork and other methods
Before getting on to the substance of these remarks, perhaps I should provide a little background information to explain the perspective from which they originate. I am both an anthropologist and a statistician (rather an unusual combination!) but I am not a demographer and nor do I have any formal training in qualitative data analysis.
There has been a tendency at this workshop to equate anthropological methods with qualitative interviewing, but I think that this equation is mistaken. Anthropologists are an individualistic and disputatious lot, so only a brave or foolish person would try to set out definitively what anthropological method is. At most one can talk of a range of methods, including that of following one's own intuition, which different anthropologists draw on in ways that suit their research problems and personalities. As such, anthropological methodology is both less and more than qualitative interviewing - less, because many (not all) anthropologists can be very free and easy in their interviewing methods, and still more in the way they analyse their notes - but more, because their impressions are based, not just on interviews, but on living with the people concerned, which includes casual chats, observation, arguments, socialising, attending rituals, noting how they decorate their houses, and eventually a level of imaginative involvement that generates ideas whose precise relation to specific bits of data one may not actually be sure of.
This may sound like a critique of anthropological data-gathering, but it is not meant to be. Loosely structured long term fieldwork seems to me to be just about the best way of getting close to the action and, above all, of identifying unexpected causal connections. Anthropological fieldwork provides rich descriptions and informed interpretations. But these interpretations do need to be disciplined. One important way of doing this is to use quantitative methods - since most interpretations, even of apparently intangible things such as emotions or religious commitment, have at least some implications for actions or objects that can be counted - and so numbers can both check existing interpretations and suggest new ideas. For instance, having formed an idea, on the basis of conversations with a fairly non-representative sample of people and some personal reflexion, that changes in spatial marriage patterns were causally connected with changing patterns of economic cooperation, it was useful to be able to check this interpretation by collecting data about the geographic origins of married couples living in the village, to sort the data by decade of marriage, and compare the results with local statistics of economic change. In this situation it is of course very important not to rig the results - and so one should use either complete enumeration or random sampling. Unfortunately many anthropologists prefer to base quantitative findings on data gathered from their own personal networks (so-called "snowball sampling" or "ethnographic sampling"), but by doing this they lose the ability to use the figures as an independent check on the conclusions they have drawn from their fieldwork experience.
Anthropology and theory
But it is not enough to check that one' s interpretations are consistent with quantitative data from one' s field site. …