Veering From a Narrow Path: The Second Decade of Social Norms Research

Adolescents and young adults, particularly young women in low- and middle-income countries, face numerous and well-documented threats to their sexual and reproductive health, including adolescent pregnancy and childbearing, much of it unintended; complications from unsafe abortions; HIV/AIDS and other sexually transmitted infections; and child marriage [1], [2], [3]. To address these challenges, governments and nongovernmental organizations have attempted a range of intervention strategies, such as comprehensive sexuality education in schools [4], youth-friendly sexual and reproductive health services [5], and cash transfers [6], [7]. In many cases, these interventions seek to improve sexual and reproductive health outcomes by changing behaviors such as sexual activity and condom and other contraceptive use, and attempt to bring about these behavioral changes through changes in individual knowledge, attitudes, beliefs, and access to services. Systematic reviews and meta-analyses suggest that such interventions frequently have positive impacts on knowledge and attitudes but less often have effects on self-reported behavioral or objectively measured biological endpoints [8].

Practitioners and investigators working on adolescent and youth sexual and reproductive health have long recognized that the behavioral and health outcomes they seek to influence are functions not merely of knowledge, beliefs, and attitudes of individual young people but also of the social contexts in which these young people are embedded. The term social norms has given people working in this area a language for talking about these contextual influences. Yet considerable confusion has been generated by the fact that the term social norms means different things to different people. Horne [9] provides a good general definition of social norms: “rules, about which there is at least some consensus, that are enforced through social sanctions” (p. 5). This definition is intentionally broad, allowing it to function as an umbrella term for a wide range of more specific types of social norms. Theorists have sometimes distinguished between norms that are essentially behavioral regularities in a community or reference group, and norms for which there exists a shared sense of moral imperative around a behavior. Even here this is no uniform terminology. Bicchieri [10] uses the term empirical expectation for norms consisting solely of behavioral regularities, and the term normative expectations for those involving social imperatives; Lapinski and Rimal [11] use the terms descriptive norms and injunctive norms for essentially the same concepts. This lack of a shared conceptualization and corresponding vocabulary is mirrored by a wide range of approaches to measuring social norms in empirical studies [12].

Within this context, the Learning Collaborative (LC) to Advance Normative Change for Adolescent Sexual and Reproductive Health and Well-Being was established as a platform for sharing and discussing emerging evidence, practices, and lessons learned among practitioners and investigators working in this area. The LC is funded by the Bill and Melinda Gates Foundation and led by the Institute for Reproductive Health at Georgetown University (convener) and FHI 360 (co-convener), has a steering committee with members from ten agencies and organizations, and has been joined by over 300 individuals from more than 100 organizations. The LC launched with a convening meeting in December 2016 and then engaged members in processes of collective learning and sharing of results through virtual meetings, webinars, and the dissemination of products through web-based libraries. The LC also initiated this special issue of the Journal of Adolescent Health as a means for sharing evidence and understanding related to normative change and adolescent and youth sexual and reproductive health (AYSRH).

While the contributions to this special issue are heterogeneous, they have a common theme: repudiation of the narrow approach to social norms that has been predominant in psychology and communication sciences on the basis that it is inadequate to address the complexities of AYSRH. At the risk of oversimplifying, the narrow approach has been to focus on behavior-specific norms and related social consequences as potential determinants of the same behavior. If the behavior of interest is condom use, for example, this approach would ask: Are individuals who perceive that condom use is more common among a reference group to which they belong, or who believe that approval of condom use is widespread among that reference group, or who think that negative social sanctions are likely for those who do not use condoms, more likely to use condoms themselves? The contribution in this special issue that comes closest to emulating that paradigm is a study of modern contraceptive use among adolescent women in Ethiopia and Tanzania [13]. Yet even this contribution departs from the paradigm in two important ways. First, rather than using individual perceptions of the prevalence of modern contraceptive use as the independent variable, the authors use the actual prevalence of use among other women in participants' communities, thereby attempting to get at the collective nature of social norms. Second, they acknowledge that these social norms may not operate independently of other factors, positing that their influence on behavior will be dampened by exposure to mass media—a hypothesis that they confirm in Ethiopia but not in Tanzania.

Two other contributions also come close to emulating the narrow approach to social norms, both with a focus on child marriage. One applies a popular approach to the identification of collective norms [10] to this topic in Malawi [14]. The other does essentially the same thing retrospectively to previously collected qualitative data from Brazil, Guatemala, and Honduras [15]. Interestingly, both find that child marriage continues to be common in these settings despite the absence of an injunctive norm or normative expectation directly supporting this practice. The authors point out that child marriage in these settings must, therefore, be driven by other factors, which may include social norms that are not specific to child marriage per se but that nevertheless impinge upon the agency of and opportunities available to adolescent women. This insight leads to the useful distinction between the direct and indirect influence of social norms. The former term refers to when social norms specific to a given behavior may influence that same behavior, whereas in the latter social norms specific to other behaviors indirectly influence the behavioral outcome of interest.

This notion that one behavior (or set of behaviors) may be influenced by social norms pertaining to other behaviors is in fact a foundational assumption for two contributions to this special issue. The first of these examines how, in one part of Ethiopia, perceived descriptive and injunctive norms regarding education, marriage, and nutrition influence adolescent women's perceived agency in matters related to sexual and reproductive health [16]. Using multiple regression analysis, they find that, net of several sociodemographic variables, five out of six of these perceived norms variables are significantly associated with perceived agency. Another contribution describes the development of new measures of social norms related to transactional sex among adolescent and young adult women in Uganda [17]. Rather than focusing on social norms regarding transactional sex itself, however, this work uses experimental vignettes to measure social norms regarding men's obligation to provide material support to women and the sexual and other obligations that this support generates for young women. The complexity and diversity of social norms that may be related to AYSRH also helps account for the broad range of measures collected by the LC's Measurement Working Group and made available through the Advancing Learning on Gender Norms Web site and its suggestion that in many cases formative research including qualitative assessment may be necessary to identify the range of relevant norms operating in a given setting [18].

A commentary in this special issue by the LC's Theory Working Group provides some understanding of the need for this complexity [19]. Many social norms, they argue, are gender specific, applying unequally to girls and boys or to women and men. Moreover, they argue gender norms and social norms more broadly exist not in a vacuum but rather as parts of a broader social system that includes formal and informal institutions and power relations at multiple levels of social organization.

The repudiation of the narrow approach to social norms is also evident in the contributions to this special issue that focus on social norms interventions. Social norms interventions from the fields of psychology and communication sciences have generally sought to influence behaviors by changing individual's perceptions of how prevalent those same behaviors are in the reference groups with which they identify or by changing their perceptions of the level of approval or disapproval of that behavior in the reference group. As documented in a recent review [20], such interventions have been used to reduce binge drinking on college campuses by correcting misperceptions about the prevalence of that behavior and to reduce households' use of electricity by providing information about the energy consumption of other households in their areas. The social norms interventions discussed in the contributions to this special issue, however, have little in common with those approaches. In a review of the scale-up of 13 social norms interventions targeting AYSRH [21] and a related commentary [22], the authors identify five practices that have been used successfully in the scale-up of these interventions. All the interventions, however, were multicomponent programs combining elements having little to do directly with social norms, such as family life education and the creation of safe spaces for youth, with activities focused on other groups such as parents and teachers as well as components such as mass media campaigns that were intended for the community as a whole. The multicomponent nature of those interventions makes it difficult if not impossible to know the extent to which their beneficial impacts may be attributable to changing social norms as opposed to, say, changing individual beliefs or attitudes or increasing adolescent women's access to sexual and reproductive health services. Relatedly, another commentary in this issue [23] argues that the interdependence between social norms (including gender norms) and the broader structures of the social and political system implies that interventions aimed at creating structural change may be needed before programs that directly target social norms can be effective.

Where does all of this leave us? The LC has served a valuable function in sparking a more integrative discussion than has existed in the past about what social norms are; how they arise, are maintained, and change; how they relate to other aspects of social organization and to purely individual-level attitudes and beliefs; what types of interventions, under what conditions, may be effective in changing social norms and improving AYSRH; and how those interventions can be scaled up. But clearly that discussion is far from over. As one commentary [22] aptly put it, this field of study and endeavor is in its own adolescence. More empirical work, more theorizing, and perhaps most importantly more discussion are needed to bring this field to a state of maturity in which its fullest potential contribution to AYSRH may be realized. Two specific steps would be helpful. First, while the phrase social norms may continue to serve as an umbrella, the field should agree upon and adhere to a clear set of terms for different types of social norms. The distinctions between descriptive and injunctive norms, between individual-level perceived norms and group-level collective norms, and between direct and indirect norms, are helpful. A set of agreed upon definitions and corresponding terminology would go a long way toward putting the field on a more solid scientific footing. Second, when it comes to multicomponent interventions seeking to change behavior in part by changing some aspect of social norms, the creators of those interventions should provide logic models that clearly delineate which intervention activities are thought to influence which types of social norms. Moreover, it would behoove impact evaluations of such interventions to conduct mediation analyses such as those described by O'Leary et al. [24]. In doing so, investigators can capitalize on an important opportunity to learn not only how efficacious interventions achieve their impacts and why ineffective interventions fail but also about how social norms influence behavior in a way that will complement observational studies.