Making and unmaking a toxic world

In recent decades, toxic risks have tremendously grown, and their regulation has become a major concern for exposed communities and states. Facing uncertainty about these emerging risks, their management has become more complex, with implications we seldom are able to foresee. Yet, many of our current concerns over toxic exposure have a long history. Toxic producing industries have often successfully obstructed scientific and regulatory processes, and many problems of industrially induced exposures have in turn been left unaddressed. Extant research shows how corporations and corporate lobbies influence policy preferences and media visibility of environmental issues, but mechanisms through which preferences of a given organization are adopted by others are rarely studied. Moreover, while the influence of corporations on the culture of political institutions is increasingly recognized, ways in which their ideas are developed and diffused in research and policy remain unclear. So far, data accessibility and secrecy around corporate practices has limited our understanding of the pathways through which industrials coalesce and cohesively act to influence environmental regulations.

In this project, I seek to answer three research questions: 1) How does corporate strategies of influence spill over between corporations? 2) How does it spill over between individual cases of toxic exposure, and 3) How do changes in policies and regulations impact corporate strategies of influence? By looking at spillover of communication strategies between corporate actors, I hope to reveal the organizational sources of environmental ideas, and how those ideas then affect environmental research and policy.

I am proud to be the first researcher to systematically use the groundbreaking Toxic Docs database. These data, hosted at the Columbia University School of Public Health, comprise millions of pages of previously classified corporate documents spanning the 20th century. They provide unprecedented insight into corporate behavior that is not distorted by public-facing strategies. Their size and time coverage (1900-2009) make them especially suitable for computational analyses. In two chapters of my dissertation, I used these data to conduct longitudinal network analyses on the cases of lead and silica.

I will exploit the full database of 8 cases and make use of word embeddings to measure linguistic convergence through cosine similarity. I will use this measure as a proxy for shifting ideas and communication strategies, and model temporal changes throughout the 20th century. Using these methods significantly broadens the temporal and cross-sectional scope in our understanding of toxic substance controversies. I will analyze both cases that are prominently studied (e.g., asbestos) and that are not often engaged with (e.g., PFAS). This will allow for insight about the shifting strategies of industries in “old” and “new” exposures. My previous experience in communicating large-N computational research has taught me that audiences generally seek in-depth examples to make sense of the broader findings. When they are not rigorously built into the research design, however, the examples provided are not reliable and do not reflect the quality of the computational work. I will systematically identify early and late adopters of spillovers, and rely on extensive archival research to contextualize some of the project’s findings.

</div> ### Hypotheses