- Knox, Dean. 'A Model for Path Data.' (with application to sectarian walking routes in Baghdad)
Path data describes the steps that an actor takes to get from point A to B. It offers researchers the opportunity to test theories about network navigation, e.g. in social and geographic networks. For example, path data can show whether individuals avoid out-group neighborhoods in their daily walking routes, resulting in societal inefficiencies and reducing inter-group contact. This data can also reveal how voters search social networks for political information, which may distort the information they ultimately receive. However, the sequential decision-making process in path data violates the underlying assumptions of existing models, which assume some form of conditional independence between observations.
I propose a new random-path model (RPM) that explicitly captures this pathwise dependence, develop an estimation procedure, and demonstrate its properties. The RPM builds on a random-walk model, incorporating a realistic but difficult-to-analyze constraint to account for the fact that actors are purposefully navigating toward a destination. I validate the model in an analysis of the U.S. Interstate Highway planning process, where existing approaches fail to recover a known qualitative benchmark.
Finally, the RPM is used to test two competing explanations of Baghdad’s recent segregation. Using smartphone-based behavioral data from Sunni and Shia participants in a field activity, I show that a need-based model of residential sorting—when families flee mixed neighborhoods to avoid political violence—is insufficient to explain participants’ walking routes alone. Instead, their choices reveal that conflict has also created significant taste-based aversion to out-groups in a city once known for its cosmopolitanism. These results suggest that societal preferences have shifted in a way that makes Baghdad’s eventual re-integration unlikely.
- Knox, Dean, Teppei Yamamoto, Matthew Baum, and Adam Berinsky. 'Design, Identification, and Sensitivity Analysis for Patient Preference Trials.' (with application to the behavioral impact of partisan media)
Social and medical scientists are often concerned that the external validity of experimental results may be compromised because of heterogeneous treatment effects. If a treatment has different effects on those who would choose to take it and those who would not, the average treatment effect estimated in a standard randomized controlled trial (RCT) may give a misleading picture of its overall impact outside of the study sample. Patient preference trials (PPTs), where participants’ preferences over treatment options are incorporated in the study design, provide a possible solution.
In this paper, we provide a systematic analysis of PPTs based on the potential outcomes framework of causal inference. We propose a general design for PPTs with multi-valued treatments, where participants state their preferred treatments and are then randomized into either a standard RCT or a self-selection condition. We derive nonparametric sharp bounds on the average causal effects among each choice- based subpopulation of participants under the proposed design. Finally, we propose a sensitivity analysis for the violation of the key ignorability assumption sufficient for identifying the target causal quantity. The proposed design and methodology are illustrated with an original study of partisan news media and its behavioral impact.
- Christia, Fotini, Dean Knox, and Jaffar Al-Rikabi. 'Networks of Sectarianism: Experimental Evidence on Access to Services in Baghdad.'
The relationship between ethnic fractionalization and lower availability of public goods and services is now treated as an empirical regularity. Using a pool of over 300 participants from paired Sunni and Shia neighborhoods in the highly sectarian context of contemporary Iraq, we conduct a novel small-world network experiment in which participants are randomly assigned to obtain information about local government services in Sunni- or Shia-dominated target areas. We trace how participants draw on their social networks and show that segregated social networks and different patterns of network search result in differential levels of access to services between groups.
Contrary to expectations, we find that the politically dominant majority Shia group is substantially less able to access public services than the minority Sunni group. They pursue an inefficient network search strategy that relies on lower-quality contacts, and are less able to leverage their social ties into costly assistance. The minority group appears to have developed better strategies for obtaining resources to which it would otherwise be denied access.
- Christia, Fotini, Elizabeth Dekeyser and Dean Knox. 'Gauging Shia Public Opinion: A Survey of Iranian and Iraqi Religious Pilgrims.'
This paper exploits the opportunity afforded by Shiite religious pilgrimage sites in southern Iraq (in Karbala and Najaf) to survey over 4,000 observant Shiites from Iran and Iraq. We explore attitudes towards their respective governments and the West, as well as opinions on sectarian conflicts raging across the Middle East in Syria and Iraq—including the rise of ISIS and Iran’s nuclear program. The Arba’een pilgrimage, which attracts several million people every year, is a unique opportunity to survey an influential—religious Shiites arguably form the backbone of support for the Iranian and Iraqi governments—yet largely understudied subpopulation of the Middle East. Despite its massive and public nature, this religious event is largely unknown to the Western world, as are the views of the observant Shiite Muslims who attend it.
On a substantive level, this project aims to fill the existing gap by characterizing the opinions of Iranian and Iraqi Shiites on a range of salient political issues at a pivotal time: The US is fighting ISIS in Iraq and Syria, has negotiated an agreement with Iran on its nuclear program, and is confronted with intense and violent intra-sectarian conflict in places like Syria, Yemen and Bahrain. Methodologically, it employs innovative survey sampling techniques for targeting hard-to-access populations, as well as survey experiments intended to elicit truthful responses to sensitive political questions.
- ● (see also Modeling > ‘A Model for Path Data’)
VIDEO AS DATA
- Knox, Dean and Christopher Lucas. 'A Model for Measuring Emotion in Political Speech.'
Though we generally assume otherwise, humans communicate using more than bags of words alone. Auditory cues convey important information, such as emotion, in many phenomena of interest to political scientists. However, in part due to the relative difficulty of processing audio data, research has disproportionately focused on the textual component of pre-transcribed corpora. We develop a hidden Markov model for emotional analysis to complement and extend existing methods for text analysis. We also make available software that implements our methods and large corpora of text for further analysis.
- Goplerud, Max, Dean Knox and Christopher Lucas. 'The Rhetoric of Debate: Conflict and Party Branding in Parliamentary Speech.'
We develop a novel theory as to why MPs engage in rhetorical conflict in parliamentary debates. We contend that doing so helps differentiate their party from their rivals and create a coherent and distinct party label. Given these incentives, we expect rhetorical conflict to vary with the public visibility of the debate and the status of the speaker.
To test this theory, we leverage recent methodological developments in the analysis of audio and video as data. With an original corpus of New Zealand parliamentary debate videos, we use the recently developed Speaker Affect Model (SAM, Knox & Lucas, n.d.) to measure (1) the mode of speech—either calm or conflictual—used by the speaker and (2) the presence of heckling in the chamber. Our results show that more prominent party representatives tend to deploy more conflictual speech and are the target of more heckling. Moreover, clashes are less frequent in technical stages of debate, where they may undermine cooperation on specific issues.
- Knox, Dean, Christopher Lucas and Justin de Benedictis-Kessner. 'Words Don't Fit the Picture: Use and Effectiveness of Sentimental Speech in the 2012 U.S. Presidential Campaign.'
Emotions play a powerful role in political communication. Building on recent advances in machine learning for emotion detection (Knox and Lucas, n.d.), we offer the first automated analysis of emotional displays in political video corpora. With this analysis, we examine politicians’ use of emotional appeals in U.S. elections and the resulting psychological effects on the electorate. We use a nested hidden Markov model for auditory and visual cues to estimate the prevalence of different emotions in presidential campaign speeches. We then describe which emotional modes of speech vary across issue areas. These descriptive findings are paired with an experiment that varies candidates’ emotional displays while maintaining the content of their speech to estimate the effect of these emotions on public perceptions. Taken together, our results demonstrate that political campaigns are sophisticated users of emotions for electoral gain.
Work in Progress
- Knox, Dean. “Identifying Peer Effects under Homophily with an Instrumental Variable: Patronage and Promotion in the Chinese Bureaucracy.”
- Knox, Dean and Christopher Lucas. “A New Method for Survey Augmentation.”
- Knox, Dean. 2012. “Nuclear Security and Nuclear Emergency Response in China.” Science & Global Security, Vol. 20, No. 1.
- Knox, Dean. 2012. “The Comprehensive Test Ban Treaty: Foundations, Context, and Outlook.” In Handbook of Nuclear Proliferation, ed. Harsh V. Pant. London: Routledge.
- Jakobson, Linda and Dean Knox. 2010. “New Foreign Policy Actors in China.” Stockholm: SIPRI. Translated as Chuugoku no Atarashii Taigai Seisaku: Darega Donoyouni Kettei Shite Irunoka. 2011. Tokyo: Iwanami Shoten.
- Jakobson, Linda, Paul Holtom, Dean Knox, and Peng Jingchao. 2011. “China’s Energy and Security Relations with Russia.” Stockholm: SIPRI.
- Knox, Dean. rpm: Manipulation and Analysis of Path Data with the Random Path Model. [under development]
- Knox, Dean and Christopher Lucas. feelR: Feature Extraction and Analysis for Emotion Detection in Audio-Visual Data. [under development]
- Lucas, Christopher, Dean Knox, Dustin Tingley, Thomas Scanlan, Shiv Sunil, Michael May, and Angela Su. transcribeR: Automated Transcription of Audio Files Through the HP IDOL API. Transcribes audio to text with the HP IDOL API. Includes functions to upload files, retrieve transcriptions, and monitor jobs.