Carlo Schwarz examines social networks to understand human actions

Carlo Schwarz, Associate Professor in the Department of Economics and research fellow of the Baffi Centre of Bocconi University, is also the principal investigator of research project entitled “CHAIN – Comprehending Human Action using Social Networks”. We have asked him some questions about the project.

When have you started this research project?

The ideas behind CHAIN grew out of my earlier work on social media and its societal effects, research I have been doing since around 2016, when my co-authors and I started investigating the link between social media and hate crime. Over the years, I became increasingly interested in a broader question: how do social networks, both online and offline, shape the decisions people make in high-stakes settings? I wrote the CHAIN proposal in 2023 and the ERC Starting Grant was awarded in 2024, so the project is now well underway.

Who is funding this research project?

The project is funded by the European Research Council through an ERC Starting Grant. I am very grateful for the support by the ERC, since the funding is critical because it allows me to build a research team, purchase proprietary datasets, and hire research assistants for the extensive data collection that each of the three projects requires.

What are the research questions?

The overarching theme of CHAIN is straightforward: we want to understand how social networks influence what people do in settings that really matter, science, policy-making, and democratic participation. The project is organized around three studies, each tackling a different piece of this puzzle.

The first project asks how social media affects the production of scientific knowledge. Specifically, we study what happens when economists join Twitter. Does it change how many papers they write, who they collaborate with, and which topics they work on? Despite how widely social media is used in academia, there was essentially no causal evidence on these questions before our study. We are particularly interested in whether social media creates a kind of herd behavior in research, where everyone ends up working on similar “hot topics”, which would be a real cost for the diversity of scientific inquiry.

The second project shifts the focus to politics. We study the role of congressional staffers in the US legislative process. Politicians tend to get all the attention, but in practice, staffers draft the bills, handle negotiations, and bring policy expertise to the table. Despite this, their influence has been very difficult to study because it is hard to separate what a politician does from what their staff does. We overcome this by exploiting the educational and professional networks that connect staffers to one another, which gives us a way to isolate the causal effect of staffer expertise on legislative outcomes.

The third project looks at the demand side of politics by studying the effects of WhatsApp in India. In 2018, WhatsApp introduced a limit on the number of times a message can be forwarded, a form of algorithmic regulation aimed at slowing the spread of misinformation. We use this policy change to study whether restricting information flows on social media actually affects political outcomes like protests, political attitudes, and election results. India is the ideal setting: it is the world’s largest democracy and has the largest WhatsApp user base.

Why this topic matters for people?

At its core, CHAIN is about understanding how the networks we are embedded in shape the information we receive and the choices we make. These are not abstract questions. If social media changes which research topics scientists pursue, that has implications for the direction of innovation. If staffers quietly steer legislation, that matters for how we think about democratic accountability. And if a simple change to a messaging app’s forwarding rules can reduce the spread of misinformation and affect election outcomes, that is directly relevant for anyone thinking about how to regulate technology platforms.

The three projects also reinforce each other. The first and second projects both highlight the importance of educational ties between individuals, PhD cohorts among economists, alumni networks among staffers. The second and third projects study opposite sides of the political process: the supply of policy through staffers and the demand for policy through citizens. And the first and third projects connect through their focus on information flows, one in science, the other in politics. Together, they paint a broader picture of how connectivity, for better or worse, shapes society.

Which data are you analysing?

Each project relies on a different combination of data sources, which is part of what makes the overall research program ambitious but also what keeps it interesting.

For the first project, we built an extensive database linking over 25,000 economists to their Twitter accounts, using information from RePEc, a large repository of economics research that also records affiliations, websites, and social media handles. We matched this with detailed records of publications, citations, and co-authorship networks. Crucially, we also collected information on economists’ PhD cohorts from the RePEc Genealogy project and from CVs, which forms the basis of our identification strategy.

For the second project on congressional staffers, we use data from commercial databases that were originally designed as a resource for lobbying contacts but turns out to be an excellent source for research. They the near universe of US congressional staffers and includes their career histories, educational backgrounds, and which congress members they worked for. We combine this with publicly available data on legislators’ voting behavior, bill sponsorship, and DW-NOMINATE scores from Congress.gov and VoteView.com, as well as data on political speeches and social media activity.

The third project on WhatsApp and Indian politics combines web traffic data from COMSCORE to measure regional WhatsApp usage, mobile internet availability data from Collins Bartholomew, conflict and protest event data from GDELT and ACLED, individual-level survey responses from the Gallup World Poll, and regional voting results from Indian government sources. The global coverage of most of these data sources also opens the door to extending the analysis beyond India in the future.

What is the state of the project at the moment?

The three projects are at different stages of completion. The first project, on the impact of Twitter usage on economic research, is the most advanced. We have assembled the full dataset, linking over 25,000 economists to their Twitter accounts, PhD cohorts, publications, citations, co-authorship networks, and research topics, and completed the empirical analysis. The paper is nearly written up and a working paper should become available soon. We have also started presenting the findings at conferences and seminars.

The second project, on congressional staffers, is also making good progress. We have completed the data collection, assembling the comprehensive database of US congressional staffers from Leadership Connect and linking it to legislators’ activities. We have prepared the data and obtained initial results. A working paper should follow in the summer.

For the third project, on WhatsApp sharing limits in India, we have run some preliminary analyses on a subset of the data. The next step is to expand the analysis to the full set of outcomes and complete the data collection for the remaining data sources. Overall, we are well on track to complete the project by the end of the grant period in 2029.

Is there any conclusion that you can share regarding your research?

Yes, particularly for the first two projects, we already have an interesting set of results. In the first project, we provide the first causal evidence on the impact of Twitter usage on economic research. We establish four main findings. First, Twitter usage raises research output, with the average user producing approximately one additional paper every five years. Second, Twitter users receive an additional four citations per year, driven entirely by citations from other Twitter users. Third, Twitter users become significantly more likely to co-author with and cite the work of other economists on Twitter. Fourth, and perhaps most importantly from the perspective of information economics, Twitter shapes the direction of research: adopters’ work becomes increasingly similar to that of other platform users and increasingly dissimilar to that of non-users. These results highlight both the benefits and potential costs of social media for the scientific process, while Twitter boosts visibility and output, it may also contribute to a convergence of research agendas among its users.

For the second project, we find that congressional staffers are a key input into the legislative production process. Using a novel identification strategy that exploits election-induced turnover of legislators combined with educational networks that mediate staffer mobility, we show that greater staffer expertise causally increases the number of bills legislators introduce in a given policy area. Moreover, staffer expertise improves legislative success: for majority-party members, higher staffer expertise increases the probability that bills pass at least one chamber and ultimately become law. Interestingly, no such effects are found for minority-party members. These findings recast policy-making as a team-based process, in which the expertise and networks of staffers play a far more important role than previously understood.

For the last, project the very preliminary results appear to suggest that the sharing limit reduced the prevalence of violent protests, but it is too early to tell if these results will hold up in the final data.