Understanding Name Generator Order Effects When Eliciting Social Networks in Rural Uganda: A Randomized Controlled Trial

Interpersonal relationships are one feature of the social environment that have been shown to have important influences on health and health behaviors (Berkman et al., 2000; Berkman & Syme, 1979) and mental health (Tsai et al., 2014). While the majority of these studies have been conducted in resource-rich settings, studies conducted in resource-limited settings have similarly demonstrated associations between social networks and health (Perkins, Subramanian, & Christakis, 2014). In this literature, the tie between two persons is the important unit of measurement. Classically, the social tie is typically elicited using a "name generator" survey that inquires about ties based on the affective content of the relationship (Laumann, 1966; Wellman, 1979) or normative bonds such as kinship ties (Kleiner & Parker, 1976).

Another popular approach to eliciting social ties has employed multiple name generators inquiring about specific types of interactions, such as discussions about important matters or exchanges of money (Burt, 1984; Marsden, 1987; McCallister & Fischer, 1978). These types of name generators have strong face validity, are less likely to be subjectively interpreted in different ways by different respondents, and are therefore likely to be more reliable. These are accompanied by a heavy respondent burden, however, and single-item name generators may perform comparably to multiple name generators only under limited circumstances (Marin & Hampton, 2007). It remains unclear, however, whether the order of name generators may affect study participant responses. While no studies have examined this empirically, there is a broad literature on public opinion polling suggesting that alternative specifications of question ordering can have important effects on respondents' answers (McFarland, 1981).