Danny Klinenberg

Why do people join domestic violent extremist organizations? This paper examines an understudied reason: organizational outreach. I study how the inflow of new members to the Oath Keepers, until recently America's largest paramilitary organization, changes when the group's leadership employs three tactics: showcasing their ideological zeal through armed standoffs with the government, membership discounts, and sports sponsorships. Using a variant of the synthetic control method, I find that standoffs increase new memberships by 150 percent, discounts increase new memberships by over 60 percent, and sports sponsorships decrease new memberships. Membership is less responsive in counties with higher income inequality, but more responsive in politically conservative counties.

Social media has become an outlet for extremists to fundraise and organize. While governments deliberate on how to regulate, some social media companies have removed creators of offensive content—deplatforming. I estimate the effects of deplatforming on revenue and viewership, using variation in the timing of removals across two video-streaming companies—YouTube, and its far-right competitor, Bitchute. Being deplatformed on YouTube results in a 30% increase in weekly Bitcoin revenue and a 50% increase in viewership on Bitchute. This increase in Bitchute activity is less than that on YouTube, meaning that deplatforming works in decreasing a content creator's overall views and revenue.

Synthetic control methods are a popular tool for measuring the effects of policy interventions on a single treated unit. In practice, researchers create a counterfactual using a linear combination of untreated units that closely mimic the treated unit. Oftentimes, creating a synthetic control is not possible due to untreated units' dynamic characteristics such as integrated processes or a time varying relationship. In this article, I investigate a new approach to estimate the synthetic control counterfactual incorporating time varying parameters to handle such situations. This is done using a state space framework and Bayesian shrinkage. The dynamics allow for a closer pretreatment fit leading to a more accurate counterfactual estimate.

At the start of the 2020 school year, some colleges chose to reopen in person while others offered primarily online classes. We find that colleges responded to financial and other incentives largely as one might expect. Larger shares of revenue attributed to in-person activities, such as dorms and dining halls, led schools to reopen in person. In general, the share of revenue due to tuition and fees had little association with reopening in-person, which is consistent with the idea that the effect of the mode of reopening on enrolment was ambiguous. However, private schools experiencing financial distress due to tuition and fees were more likely to reopen in-person while public schools were less likely.

We use response curves in a repeated game to analyze key aspects of mutual deterrence: escalation, de-escalation, incomplete deterrence, and deterrence by denial. In this approach, episodes of violence are due to interacting response curves, which disincentivize opponents from attacking through both deterrence and compellence. Both sides punish attacks to maintain credibility in future episodes, disincentivizing larger attacks and yielding nonviolent lulls. We empirically estimate those curves using detailed incident data from the Israel-Gaza conflict between 2007 and 2017. Our estimates match the dynamics of the raw data: very frequent episodes of low lethality violent exchange. Response curves are stable and exhibit a posture consistent with incomplete deterrence: i.e., episodes de-escalate, but to a violent equilibrium. Israeli missile defense shifts the Gazan response curve to a less aggressive posture, as predicted by theory.

Many applied economic studies aim to estimate strategic behavior through reaction curves. Examples include two-sided conflicts, economic trade wars, and algorithmic pricing between firms. Analysis is usually performed at a pre-specified time interval, such as days, weeks, months, or years, using a Vector Autoregression (VAR). Yet sides may respond within a day to one action, but wait a month after another. If data is recorded in arbitrary time intervals, then the researcher may mistake waiting to act for inaction. We analytically show that VAR analyses do not recover true reaction curves if the timing of reaction is not accurately recorded. We discuss alternative approaches and investigate their usefulness in a Monte Carlo simulation.