Carlos Cinelli
(Department of Statistics, University of Washington)

Transparent and Robust Causal Inference in the Social and Health Sciences
28 October 2021
14.00-15.30 

The past few decades have witnessed rapid and unprecedented theoretical progress on the science of causal inference, ranging from the “credibility revolution” with the popularization of quasi-experimental designs, to the development of a complete solution to non-parametric identification with causal graphical models. Most of this theoretical progress, however, relies on strong, exact assumptions, such as the absence of unobserved common causes, or the absence of certain direct effects. Unfortunately, more often than not these assumptions are very hard to defend in practice. This leads to two undesirable consequences for applied quantitative work: (i) important research questions may be neglected, simply because they do not exactly match the requirements of current methods; or, (ii) researchers may succumb to making the required “identification assumptions” simply to justify the use of available methods, but not because these assumptions are truly believed (or understood).  In this talk, I will discuss new theory, methods, and software for permitting causal inferences under more flexible and realistic settings. In particular, I will present a novel suite of sensitivity analysis tools for identification via regression adjustment and instrumental variables, which can be immediately put to use to improve the robustness and transparency of current applied research. I will also show graphical tools for the algorithmic derivation of sensitivity curves in arbitrary linear structural equation models. These tools empower scientists, and policymakers to both examine the sensitivity of causal inferences to violations of its underlying assumptions, and also to draw robust and trustworthy conclusions from settings in which traditional methods fail.

Register here for this meeting! 

Dr Niclas Moneke 
(Department of Economics, University of Oxford)

Professor Mario Luis Small
(Grafstein Family Professor of Sociology, Department of Sociology, Harvard University)

 Lent Term 2020/21

What is “urban data justice”?: Defining, conceptualizing, and exploring data use, re-use, and refusal for racial justice
Speaker: Dr Matthew Bui
(Provost’s Postdoctoral Fellow, The NYU Alliance for Public Interest Technology)
Date: 21 January 2021
Abstract: This talk explores the relationship between marginalized communities of color and data, foregrounding questions about power, inequality, and justice within the aspiring “smart” city of Los Angeles. First, I will briefly touch on a study that proposes a typology of community-based engagements with data for racial justice: namely, data use, re-use, refusal, and production. Building on this work and considering the politics of data re-use and refusal to keep powerful actors accountable, I will discuss in detail a second study, which leverages an original Yelp dataset to construct and deconstruct a novel indicator of urban displacement, using coffee shops as a flashpoint for urban change. This work theorizes and conceptualizes “urban data justice” as a community-engaged vision and praxis while also articulating and exploring a more politically engaged, grounded, and mixed-method approach to “data-driven” research and policy. In all, I ask: how do we tell stories with—and about—data? Who benefits from these dominant narratives about data? How can we subvert unequal power relations within—and of—data within an age of datafication?

Designing deliberation: Three ethnographic challenges
Speaker: Dr Nicole Curato
(Associate Professor, University of Canberra)
Date: 4 February 2021
Abstract: There have been increasing calls to implement deliberative democratic processes or minipublics to deepen citizen participation in policymaking. Advocates argue that these forums’ design features of (1) inclusion via random selection, (2) open-minded deliberation, and (3) consequential recommendations are critical in revitalising democracy today.

This presentation critically interrogates these processes’ core design features based on autoethnographic fieldwork on deliberative forums in conflict zones in the Philippines. It poses three ethnographic challenges on the ethics and politics of implementing these design features in contexts defined by fear, emotional trauma, and micropolitics of everyday life. By posing these challenges, this presentation aims to pluralise what it means to conduct ‘good’ deliberation based on normative theory grounded on ethnography.

An Imperfect Match? Gender and Racial Discrimination in Hiring Across Skill Matching
Speaker: Dr Kate Weisshaar
(Assistant Professor, Sociology Department at the University of North Carolina - Chapel Hill and faculty fellow at the Carolina Population Center)
Date: 11 February 2021
Watch a recording of this seminar here.
Abstract: This talk considers how gender and racial discrimination in hiring screening decisions is related to applicants’ match or mismatch on skills required for the job. In the talk, I will present findings from an ongoing collaboration with Koji Chavez and Tania Cabello-Hutt. Theoretically, this project works toward extending and integrating hiring matching theories with discrimination theories. We differentiate between competing theoretical predictions as to whether gender and racial discrimination in hiring is responsive or resistant to applicants’ levels of skill match for the position. To test these predictions, I present results from two original experimental studies. The first, a survey experiment, assesses whether respondents’ levels of stereotyping against Black and women applicants depend on whether fictitious job applicants’ skills match or do not match the job’s requirements. The second empirical study draws from an audit correspondence study of employers. In this study, we examine how employer bias varies across levels of applicants’ skill matching as well as the types of skills the applicants match or mismatch on. Overall, we find that increased skill matching is associated with decreased hiring discrimination, but that there is some variation in these effects across different types of skills. The talk concludes by suggesting methodological implications for future research in terms of identifying and interpreting skills, as well as research more generally on hiring discrimination.

Citation inequities in the social sciences: The case of Communication studies
Speaker: Dr Deen Freelon
(Associate Professor, UNC Hussman School of Journalism and Media)
Date: 18 February 2021
Watch a recording of this seminar here.
Abstract: Calls for equity across categories of race, gender, and national identity are nothing new in politics or the academy. Yet the social sciences continue to marginalize both research on and by members of such identity categories. To quantify these inequities, I analyze citation patterns from ten prominent journals in the field of Communication between 2000 and 2019.

The data come from Web of Knowledge, and the analysis focuses on each author’s identity and professional characteristics, including race, gender, country of employment, and discipline. As this research is currently in progress, the talk will focus on the methods used to collect and preprocess the data. Also, a preliminary descriptive analysis of gender disparities in Communication citation practices will be presented. This research offers both a model methodology and an empirical baseline for measuring inequities in citation practices across disciplines.

Using eye movements to predict large-scale voting decisions in direct democracy elections
Speaker: Dr Jason Coronel
(Assistant Professor, School of Communication, Ohio State University)
Date: 11 March 2021
Watch a recording of this seminar here.
Abstract: Over 100 countries allow people to vote directly on policies in direct democracy elections. Politicians are often responsible for writing ballot language, and voters frequently encounter ballot measures that are difficult to understand. We examine whether eye movements from a small group of individuals can predict the consequences of ballot language on large-scale voting decisions.

Across two preregistered studies (n = 120 registered voters and n = 120 registered voters), we monitored laboratory participants’ eye movements as they read real ballot measures. We found that eye movement responses associated with difficulties in language comprehension predicted aggregate voting decisions to abstain and vote against ballot measures in U.S. elections (total number of votes cast = 137,661,232). These findings expose the concerns of direct democracy elections as politicians and interest groups may inadvertently or deliberately influence election outcomes by crafting difficult-to-understand ballot language. However, our study also lays the groundwork for how these concerns can be addressed through eye movement monitoring. Since eye movements provide a continuous measure of reading performance, they can potentially reveal whether the challenges in understanding ballot language occur at the level of specific words, sentences, or the entire text. Eye movements may be able to assist researchers and policymakers in crafting ballot language that is comprehensible to a larger group of voters.

Mobilizing Social Capital for Pretrial Release
Speaker: Professor Sandra Susan Smith
(Daniel and Florence Guggenheim Professor of Criminal Justice, Harvard Kennedy School)
Date: 25 March 2021
Watch a recording of this seminar here.
Abstract: Spending more than one day in pretrial detention can significantly and negatively shape individuals’ short- and long-term outcomes. Pretrial detention places detainees at risk of serious physical harm, negatively affects case disposition, increases the likelihood of future penal system involvement, and reduces individuals’ access to a broad range of opportunities. To date, however, we know relatively little about the resources that individuals deploy to navigate the pretrial process so as to avoid some of its more negative effects. In this study we draw from 191 in-depth interviews with participants in San Francisco’s Pretrial Diversion Project to examine whether, when, among whom, and for what purposes people draw on their social networks to get out of jail, especially during the all-important first few days of detention when the collateral consequences of detention are arguably set in motion. Preliminary analysis reveals that pretrial detainees rely heavily on family members, especially mothers, to mitigate the material, social, and emotional consequences of pretrial detention, as well as for help securing release. Fellow inmates also prove to be an extraordinary resource, acting as sources of social and emotional support more often than romantic partners, friends, or jail officials. Patterns of social capital mobilization, however, vary by detainees' age, gender, race, and arrest history. These findings provide important contributions to the growing literature on the front end of criminal case processing, which has yet to systematically consider the extent to which and how social networks and social capital shape the pretrial experience and related outcomes.

Michaelmas Term 2019/20

What do we mean by a "hard-to-reach" population?: Legitimacy versus precarity as barriers to access
Speaker: Rachel Ellis, PhD 
(Department of Criminology & Criminal Justice, University of Maryland)
Date: 8 October 2020
Abstract: A myriad of articles and textbooks advise qualitative researchers on accessing “hard-to-reach” or “hidden” populations. In this presentation, I compare two studies that I conducted among justice-involved women in the U.S.: a yearlong ethnography inside a state women’s prison and an interview study with formerly incarcerated women. Although these two populations are interconnected – and both deemed “hard-to-reach” – the barriers to access differed. In the prison ethnography, “hard-to-reach” reflected an issue of institutional legitimacy, in which researchers must present themselves and their proposed study as legitimate and worthy to organizational gatekeepers. In the reentry interview study, “hard-to-reach” reflected an issue of structural precarity, in which researchers must navigate the everyday vulnerabilities of research participants’ social position to ensure inclusion and accessibility. Juxtaposing these two experiences, I propose greater nuance to the term “hard-to-reach,” such that researchers may proactively address potential institutional and structural barriers to access.

Citizen assemblies, inequality, and deliberation in rural India: An approach combining causal inference, qualitative analysis, and machine learning
Speaker: Vijayendra Rao
(Development Research Group, The World Bank)
Date: 22 October 2020
Watch a recording of this seminar here.
Abstract: This talk will review two projects that analyze transcripts from a total of 400 village meetings (gram sabhas) in rural South India. 300 transcripts (collected in 2002-4) are analyzed using “hand-made” qualitative analysis in the context of a natural experiment, and 100 (collected in 2014) are analyzed using text-as-data methods with a combination of random assignment (of women leaders) and regression discontinuity (assigning the entry of women’s self-help groups).  The results show that despite high levels of caste and wealth inequality, citizens are active participants in village meetings, that there is a great deal of gender inequality in deliberation, and that reserving seats for women presidents sharply reduces gender inequality, as does the introduction of women’s self-help groups, but they crowd out more organic forms of deliberation.

Factors associated with COVID-19 related mortality using the OpenSAFELY platform
Speaker: Elizabeth Williamson
Associate Professor of Biostatistical Methodology, Medical Statistics Department, London School of Hygiene & Tropical Medicine)
Date: 12 November 2020
Watch a recording of this seminar here.
Abstract: This talk will present an overview of the recently developed OpenSAFELY platform, which brings together primary care data and data on COVID-19 outcomes, including test results, hospital admission and mortality. 

OpenSAFELY uses a new model for enhanced security and timely access to data: we don’t transport large volumes of potentially disclosive pseudonymised patient data outside of the secure environments managed by the electronic health record software company; instead, trusted analysts can run large scale computation across near real-time pseudonymised patient records inside the data centre of the electronic health records software company.

The talk will discuss the structure of the platform and then present the recently published results about factors associated with COVID-19 related mortality and current extensions underway to develop risk prediction models allowing for the changing burden of infection over time. It will touch on other questions that we have addressed using the platform, including effects of inhaled corticosteroids on COVID-19 outcomes.

COVID Realities: A collaborative and participatory approach to understanding the experiences of low-income families during COVID-19
Speakers: Dr Maddy Power and Dr Ruth Patrick
(University of York)
Date: 26 November 2020
Watch a recording of this seminar here.
Abstract: In the early days after the outbreak of Coronavirus in the UK, there was much talk of us ‘all being in it together’, the idea that the virus did not discriminate and could affect us all. But a narrative of unity and a commonality of experience contrasts with experiences and risks of Coronavirus, and the subsequent economic shocks, which are profoundly unequal.

This talk will provide an overview of the COVID Realities project, a research collaboration between parents and carers, the Universities of York and Birmingham, and our third sector partner, the Child Poverty Action Group (funded by the Nuffield Foundation as part of their rapid-response to COVID-19). It will outline our collaborative approach to research during COVID-19, including a synthesis of existing and ongoing research into poverty in the UK (focusing on the impact of COVID-19); and the facilitation of conversations and resources for the research community on methodological and ethical challenges during a time of COVID-19.

It will describe our approach to participatory research with parents and carers who are living on a low income and explain how we strive to value varied forms of expertise within the project, drawing upon this at every stage. It will discuss the COVID Realities website, through which participants can document their experiences via online diaries, responding to a pre-recorded audio-visual ‘big question of the week’ and participating in monthly online 'big ideas' groups. Finally, it will consider some of the methodological challenges we have faced so far and outline our plans going forward.

Border-Free ELF: Using Individualized Spatial Data to Measure Ethnic Segregation
Speaker: Dr Neelanjan Sircar
(Assistant Professor, Ashoka University and Senior Visiting Fellow, Centre for Policy Research)
Date: 10 December 2020
Watch a recording of this seminar here.
Abstract: One of core problems in the spatial social sciences is the measurement of racial or ethnic segregation and fractionalization. Traditional measures, like dissimilarity (D) or Herfindahl-Hirschman (HHI) indices, are highly sensitive to the borders of the geography being used (e.g., county, neighborhood, census tract). To alleviate these concerns, we develop a mathematical framework to measure segregation using ethnolinguistic fractionalization (ELF) indices that can be calculated purely from spatial or social distances between individuals or households without the artificiality of geographic borders — what we call border-free ELF.

We apply our insights to a dataset comprised of 4380 geocoded households in the National Capital Region (Delhi metropolitan area) in India. We make three key claims. First, the use of geographic borders can significantly overstate the amount of segregation. Second, unlike traditional measures, border-free ELF can detect individual “perceptions” of spatial similarity and dissimilarity and plausibly be used to analyze the correlates of different perceptions. Finally, border-free ELF can be deployed with survey data to provide an individualized picture of the relationship between racial/ethnic fractionalization and public goods provision.