From Lazarsfeld to Data Science: Elihu Katz on the Persistent Relevance of the Two-Step Flow

Impressively, the Monday keynote at ICA 2018 is by Elihu Katz, whose considerable impact on communication research does of course reach back to the 1950s. He begins by noting the important role that Paul Lazarsfeld played in restoring interpersonal communication to the study of communication, a development which is crucial to the study of social networks today.

Lazarsfeld became interested in radio in the 1930s, and was also intrigued by the psychology of decision-making; he combined this in his studies of voters in Ohio over an extended period of time. This enabled him to identify voters who changed their minds during the course of an election campaign – a change which was attributed not mainly to media coverage, but to the role of better-informed opinion leaders. This was the basis for the theory of the two-step flow, which was be proven in subsequent studies that examined the roles of both influencers and influencees and identified different spheres of influence.

Such studies showed that in the 1940s husbands influenced wives in political matters, that young women influenced other women in fashion choices, and so on. At the time, participants were asked to provide the names and addresses of their interlocutors, and a two-step flow network could be constructed from that information. This resulted in the forced marriage of survey research and sociometry, in order to explore where this network would reach, but this connection was difficult: sociometry is based on personal ties between the members of a group, while survey research draws only on a representative sample of otherwise potentially disconnected participants.

Potential solutions were to conduct sociometry around selected survey respondents, and to study the diffusion of innovation (for instance, the adoption of new drugs) across social networks. This also drew on underlying medicine prescription data, and confirmed that sociometrically popular doctors were also the earliest adopters, that some less popular doctors were acting as innovation scouts, and that cliques adopted innovations together. A classic S-shape innovation adoption curve that pointed to tipping points in adoption also emerged. Such adoption processes could also be translated to institutional adoption processes.

Further research found that everybody finds their place in a network of communication; that these networks are relatively stable; that change flows through these networks; that some members are more influential than others; that these leaders are more exposed to the media; and that hearing opposing views is not conducive to political participation (it leads to a withdrawal from participation). But the concept of opinion leaders also needed downsizing; not everybody is part of a network; influencers and influencees are interchangeable; two steps are probably too few; influencers and innovators may be different people; influence comes from equals, not superiors; influence comes in stages and has input from different media; and influencers may also lead in resisting change.

Such research also echoes in important ways the work of Gabriel Tarde, half a century earlier. He took a similar networked view of communication, distinguishing between crowds and publics and highlighting the importance of public conversation. He believed that opinions are formed through the day-to-day exchanges of casual conversations, but such ideas had disappeared from the scholarship by the time of Lazarsfeld’s and Katz’s work; they were only subsequently rediscovered and incorporated into our understanding of communication flows – even if Tarde’s views were informed by the study of Parisian salon society rather than Ohioan housewives.

Tarde’s views have now gradually been reintegrated into communication research, and translated to more recent media environments. Tests of Tarde’s propositions about the interplay’s between mainstream media and public opinion formation found that reading newspapers increased political understanding, but that readers’ own opinions were not ideologically consistent: the more enlightened citizens should be more ideologically consistent after exposure to the media, but were found not to be.

How do new media affect such processes, then? Sandra Gonzalez-Bailon has revitalised the Tardean approach, and the work of Lazarsfeld and Katz that builds on it, by restating his propositions in the form of computational data science. Digital technologies have shifted our interest to social media networks, but many of the questions from previous studies translate to this context; here, too, here is a kind of marriage between sociometry and survey research, but building on different and much larger datasets. But the focus here is more on hashtags and topics than on opinions, but there are now important steps being taken towards a reintegration of communication research and public opinion research, which had been curiously separated in the meantime.

Protests and revolutions – features of the current politics of constant complaint – have been an important focus of this research; major protests appear to be irresistible to big data researchers, and the highlight the roles played by activists and their audiences. This enables us to study that elusive phenomenon of collective behaviour, especially in social movements. Lazarsfeld and Tarde deserve a share of the credit for the development of this field of research.

Collectively, then, we have three eras in which the two-step flow has been examined: the era of newspapers, the era of broadcast, and the era of new media. We cannot imagine how much newspapers have connected and unified the conversations of individuals, even if those individuals did not themselves read newspapers – but we can perhaps examine this for the current media environment.