Site Sampling and Snowball Sampling - Methodology for Accessing Hard-to-reach Populations

Communities and networks of refugees, migrants, the poor, hidden, powerless, and the otherwise marginalized groups, have always been of interest to social researchers, and social workers, because it is within such communities and networks that interrelated, even tangled, pathologies can be found. In service of multiple disciplines, and across a wide spectrum of social research, great effort has been extended to find members of such groups, with the aim not only of scientific curiosity but also of ameliorating their oppressive life-circumstances, and discovering the social structures in which they are embedded. The objective of this note is to call attention to a possibly-useful methodology that was long-ago applied to a study of community leadership in poor communities (TenHouten et al., 1971), which was used in conjunction with another method, snowball sampling, that has enjoyed wide application. These two methods, site sampling and snowball sampling, we propose, comprise a natural combination for acquiring data from individuals associated with hard-to-reach populations. In this study, one-criterion snowball sampling was used, but multiple-criterion sampling could also be used (TenHouten, 1992).

The most used of ‘traditional’ methodologies for reaching any population or subpopulation of individuals is residential sampling. But in trying to access those who are hidden or hard-to-reach, such research design is typically ineffective, because the individuals belonging to the targeted population are apt to not be associated with an specific residence, or, even if they are, to be only transient occupants, largely living on the ‘streets’ of large cities, or in insular communities not welcoming of strangers. Not only that, but the patterns of behavior of interest to researchers typically do not take place within residences, but outdoors.

With respect to opinion, ideology, and leadership, another traditional method has been ‘reputational’ sampling. Reputational sampling accesses individual from the ‘top’ of a community’s social hierarchy, assuming that these highly visible designated persons are indeed leaders, that that studying them will provide knowledge of community leadership structure. The ‘reputational’ technique systematically fails in efforts to uncovered the social structuration of hard-to-reach populations, almost by definition. As one dated example, Hunter (1953; see also Jennings 1964) had the heads of civic organizations provide lists of top leaders, and then had a panel of fourteen judges select the leaders from the resulting corpus of names. This method, however provides no rationale for selecting organization heads, and leads to no statistical evaluation of whether a leadership structure even exists. This method provides no way to know if the informal level of community is even known to these putative ‘leaders’, much less known in depth.

Confronted with the task of designing a sampling methodology to study the ideology and leadership structure in poor black and Latino communities following the Watts Riot in Los Angeles, the author developed, and put into practice, a methodology called site sampling, used in conjunction with snowball sampling.i This method held that:

If individual were not accessible in residential (R) setting, then they must be found in commercial or public-use areas (C) such as parks, railway stations, churches, mosques;

The rationale for these presumptions is that individuals can be located only where they are, and when they are there.

Cross-classification of these three criteria led to eight sites:

Commercial-public Inside Day (CID), e.g. the barbershop, the mall, the store, the day-care center;

Commercial Inside Night (CIN), e.g. the pub, the café, the brothel;

Commercial Outside Day (COD), e.g. the corner drug trade;

Commercial Outside Night (CON), e.g. outdoor cafes, street corners, alleys;

Residential Inside Day (RID), e.g. the apartment, the flat, the home, the gym;

Residential Inside Night (RIN), e.g. the apartment, the flat, the home, the homeless encampment;

Residential Outside Day (ROD), e.g. the porch, the commons;

Residential Outside Night (RON), e.g. the barbecue, the housing project courtyard.

To estimate population characteristics from such sites, it is necessary to estimate:

the number of persons at each site;

In an application of this method, it was found that persons located in commercial areas, who are outside of residences at night, report that they were only infrequently available in the RID and RIN sites; no amount of call-backs were able to locate many of these individuals. The RID and RIN interviews tended to generate respondents who, relative to those located at other sites, were not involved in informal community leadership, which was most concentrated in the CIN sites. There were numerous significant between-site differences in political ideology, political involvement, racial attitudes, and social behaviors.