Jeffrey L. Jensen’s research while affiliated with Abu Dhabi University and other places

We explain and document state-level fiscal developments in American Southern states from 1820–1910, focusing on their main source of revenue, progressive property taxes borne primarily by economic elites. The fourteen states in our analysis were characterized by severe economic exploitation of the enslaved and later politically repressed African-descended population by a small rural elite, who dominated the region both politically and economically. While rural elites are thought to be especially resistant to taxation, we offer a set of conditions that explains the emergence of progressive taxation and provides a coherent account of the fiscal development of these states over this period. Using an original, archival data set of annual tax revenues and select expenditure items, we show that the economic interests of these rural elites and the extent of their formal (over)representation played a critical role in shaping the observed fiscal patterns within and across these states over this period. This title is also available as Open Access on Cambridge Core.

With the growing evidence that the content on Fox News Channel (FNC) influences its viewers' attitudes and behaviors, it is important that we better understand the real-world factors driving demand for FNC, especially among those who are not already frequent viewers. Yet, our ability to identify these factors is plagued by difficult methodological issues, most notably, the selection problem of inferring the determinants of demand from content. This study overcomes these challenges by exploiting the substantial spatial and temporal variation in the Black Lives Matter (BLM) protests during 2020, the year of George Floyd's murder and other prominent police killings of Black Americans, to examine whether perceived racial threats increased FNC viewership. We construct a census block group-week panel of BLM protests and FNC ratings during 2020. With this spatial panel, we test whether the local occurrence of a BLM protest increased the portion of the local population watching FNC in the subsequent period. We find that the number of FNC viewers rose by approximately 2% following a BLM protest. At the same time, we find evidence that viewership of CNN, a moderate news channel, was unaffected and possibly declined. Our results, which survive robustness checks, tests of the modeling assumptions, and replication at the census tract level, can be interpreted as causal under reasonable assumptions. The findings provide real-world evidence that perceived racial threats increase demand for conservative media.

Scholars and policymakers are concerned about the effects of misleading information found on politically conservative social media. Yet, the ability of scholars to causally identify its effects are limited in part by two methodological hurdles when using observational data. First, few studies have a credible strategy for overcoming the issue of selection bias, whereby conservatives are likelier to consume conservative social media. Second, it has been difficult to reliably link social media activity and offline outcomes. To overcome the first problem, our identification strategy exploits quasi-random spatial and temporal differences in conservative social media usage in the lead-up to the 2020 US presidential election. Specifically, the winner-takes-all rules for awarding most states’ electoral votes means that campaigns focus their mobilization efforts only in swing states. This temporarily increased relative political interest - and conservative social media usage - in these states. Second, we use the public release of metadata from Parler, a conservative alternative to Twitter, to reliably identify the geo-location of conservative social media usage across the US during 2020. Using a county-month panel, we show that Parler activity increased more in counties in swing states in the lead-up to the presidential election. We use this as an instrument to estimate a two-stage least-squares model, which shows that greater county Parler activity over 2020 caused more COVID-19 deaths. The findings are robust to the inclusion of county-month viewership ratings of Fox News Channel. Our study demonstrates causally that politically conservative social media has negative consequences for offline outcomes.

Computational methods have become widespread in the social sciences, but probabilistic language models remain relatively underused. We introduce language models to a general social science readership. First, we offer an accessible explanation of language models, detailing how they estimate the probability of a piece of language, such as a word or sentence, on the basis of the linguistic context. Second, we apply language models in an illustrative analysis to demonstrate the mechanics of using these models in social science research. The example application uses language models to classify names in a large administrative database; the classifications are then used to measure a sociologically important phenomenon: the spatial variation of religiosity. This application highlights several advantages of language models, including their effectiveness in classifying text that contains variation around the base structures, as is often the case with localized naming conventions and dialects. We conclude by discussing language models’ potential to contribute to sociological research beyond classification through their ability to generate language.

Scholars have long tried to understand the conditions under which actors choose to use violent versus non-violent means to settle disputes, and many argue that violence is more likely in weakly-institutionalized settings. Yet, there is little evidence showing that increases in state capacity lowers the use of violent informal institutions to resolve disputes. Utilizing a novel dataset of violence—specifically, duels—across American states in the 19th century, we use the spread of federal post offices as an identification strategy to investigate the importance of state capacity for the incidence of violent dispute resolution. We find that post office density is a strong, consistent, and negative predictor of dueling behavior. Our evidence contributes to a burgeoning literature on the importance of state capacity for development outcomes.

Social capital has been shown to positively influence a multitude of economic, political, and social outcomes. Yet the factors that affect long-run social capital formation remain poorly understood. Recent evidence suggests that early state formation, especially investments in state capacity, are positively associated with higher levels of contemporary social capital and other prosocial attitudes. The channels by which early state capacity leads to greater social capital over time are even less understood. We contribute to both questions using the spatial and temporal expansion of the US postal network during the 19th century. We first show that county-level variation in post office density is highly correlated with a bevy of historical and contemporary indicators of social capital (e.g., associational memberships, civic participation, health, and crime). This finding holds even when controlling for historical measures of development and contemporary measures of income, inequality, poverty, education, and race. Second, we provide evidence of an informational mechanism by which this early investment in infrastructural capacity affected long-run social capital formation. Namely, we demonstrate that the expansion of the postal network in the 19th century strongly predicts the historical and contemporary location of local newspapers, which were the primary mode of impersonal information transmission during this period. Our evidence sheds light on the role of the state in both the origins of social capital and the channels by which it persists. Our findings also suggest that the consequences of the ongoing decline in local newspapers will negatively affect social capital.

The ratification of constitutional changes via referendum is an important mechanism for constraining the influence of elites, particularly when representative institutions are captured. While this electoral device is commonly employed cross-nationally, its use is far from universal. We investigate the uneven adoption of mandatory referendums by examining the divergence between Northern and Southern U.S. states in the post-independence period. We first explore why states in both regions adopted constitutional conventions as the primary mechanism for making revisions to fundamental law, but why only Northern states adopted the additional requirement of ratifying via referendum. We argue that due to distortions in state-level representation, Southern elites adopted the discretionary referendum as a mechanism to bypass the statewide electorate when issues divided voters along slave-dependency lines. We demonstrate the link between biases to apportionment and opposition to mandatory referendums using a novel data set of roll calls from various Southern state conventions, including during the secession crisis of 1861.