2015-2016 Winter Colloquia

ABSTRACT: A key goal of Internet.org (a Facebook initiative) is to connect the world by developing market and technological solutions to bring people online who otherwise would not be able or interested in using the Internet. In order to be effective and measure effectiveness, it is imperative for Facebook to estimate and track the number of active Internet users in countries around the world, as well as understand who the unconnected population is and what connectivity barriers they face. This talk will discuss three survey research efforts by Facebook to understand the unconnected population and the issues we face with each:
1. Extensive in-person surveys in developing countries around the globe to get precise measures of attitudes and behaviors related to Internet use. These high-cost surveys provide rich sources of information about each country and allows for a comparative examination of barriers across countries.
2. Comparison of IVR and in-person interviewing in Colombia to see if we can adequately supplement our in-person survey when new issues and questions arise.
3. A pilot study to estimate the number of Internet users from a sample of Facebook-only users using network reporting methods. This is based on the idea that survey respondents have useful information about other people whom they are connected to in their personal networks. The goal of this study is to find a cheaper, faster way to measure changing trends in Internet penetration between waves of the more costly in-person interviews.In addition, I will leave some time in the end to field questions about what it is like to work in industry and about specific opportunities at Facebook.

ABSTRACT: Big social data, such as that produced by Facebook and Twitter, have the potential to transform the social sciences and lead to advances in understanding human behavior. At the same time, novel large-scale methods and forms of collaborationbetween academia and industry raise new and important ethical and methodologicalquestions.In this talk I will discuss the Facebook Emotion study and step through several aspects ofthe study that involve important ethical decision points, and provide some insights onwhy the study generated such massive attention and criticism. I will also touch on the experience of an Internet-scale controversy, from the personal costs to the gift of criticism, and the potential opportunities to move the discussion forward.

ABSTRACT: The pre-election polls significantly under-estimated the size of the Conservative lead over Labour at the 2015 U.K. General Election, with nearly all the final polls calling the ‘horse race’ as a dead heat. The failure to predict the Conservative lead in the vote share has led many commentators to question the value and robustness of polling methodology, with some prominent figures even calling for polls to be banned in the final weeks preceding an election. In response to these events, the British Polling Council (BCP) and the Market Research Society instigated an independent inquiry into the causes of the discrepancy between the final polls and the election result, chaired by Professor Patrick Sturgis of the University of Southampton. In this talk, Mario Callegaro, member of the BPC task force, will present their findings and preliminary conclusions about the factors which led to the polling miss.A full written report is expected to be available end of March.

ABSTRACT: It has been almost twenty years since the last national study on the political homogeneity of US marital pairs. In this paper, we map the political similarity of spouses and their offspring in the era of mass polarization using a YouGov data set with a unique sampling design, and compare our results to baseline couples-offspring data from 1965. We find that spousal and inter-generational political agreement has increased dramatically, especially on policy preferences, but also with regard to partisan identification and ideology. Next, we leverage access to a complete voter file of 40,000,000 cases to validate our estimate of spousal homogeneity in party identification. We further exploit the voter file to show that political homogeneity among married couples is more a result of mate selection than a byproduct of exposure to a politically homogeneous context. Finally, we derive an observational estimate of the relative effects of mate selection and persuasion over the course of marriage on spousal political homogeneity. This analysis suggests that persuasion contributes only modestly to spousal agreement, with most of the persuasion effects occurring among female spouses. In closing, we consider the implications of mass polarization for processes of family socialization.

ABSTRACT: Generalized trust differs from other forms of specific trust in people, since it does not rely on evaluations of individuals you already have met (e.g., your family, friends, and co-workers). Instead, generalized trust refers to a general perception of unknown people (Rosenberg 1956; Uslaner 2002). This trust is, in turn, thought to help overcome collective action problems (i.e., fostering cooperation towards a mutual benefit despite the prospect of some individuals’ free riding) and has thus been proposed to be the social lubricant that makes our democracy work (Putnam 1993; 2000; Rothstein and Uslaner 2005).

In addition, generalized trust has been argued to affect or be affected by almost everything, from individual decisions of joining local charity organizations(Sønderskov 2011), to the degree of corruption in national and local governments (Rothstein and Stolle 2008; Rothstein 2011). The common ground for most of these studies is that they rely on a survey measurement of generalized trust. In fact, it is hard to find a study investigating this trust without using a survey question to measure it.

Despite this, only a few studies discuss the sources of measurement error stemming from the survey response process in the generalized trust question (some exceptions are Glaeser et al. 2000; Reeskens and Hooghe 2008; Dinesen 2011; Uslaner 2012a).This lack of research is surprising given the likelihood that survey research, at least for a foreseeable future, will not decline and that research continues to measure generalized trust mainly using survey questions.

Hence, the aim and focus of this dissertation is to evaluate sources of measurement error in the generalized trust survey question at the individual level, using survey methodology theory. Specifically, the dissertation tackles three research questions relating to survey measurement error: (1) Can the generalized trust survey question be improved by applying insights from survey methodology theories? (2) Does asking the same survey question over time, using a survey panel design, increase measurement error? (3) Building upon previous findings that ethnic diversity correlates with lower levels of generalized trust, the thesis asks whether effects of ethnic diversity on generalized trust are biased by measurement error in respondents’ answers to factual questions about the immigrant population. Lastly, using the validated generalized trust measure from the above queries, the final question asks more substantively (4) whether experiences at the individual level can affect generalized trust.

ABSTRACT: As the number of first generation immigrants and their second-generation children rises in America to the highest levels since the early twentieth century, questions about the potential political influence of these groups are of increasing interest. This study examines the effects of priming generational status on first and second-generation Americans, on the strength of ethnic and American identities, and immigration policy attitudes. We find that while priming does not appear to have a direct effect on strength of ethnic or American national identity for first and second-generation Americans, it does shift expressed attitudes toward immigration policy in a conservative, less pro-immigrant direction, for both of these groups. The implications of these initial findings and suggestions for further study are discussed.

ABSTRACT: The American public’s approval of the U.S. Congress in at an all-time low. This dissatisfaction might be a function of perceived gridlock in Congress(which might be called “volume disapproval”) and might be a function of unhappiness with the nature of the laws that have been passed (“outcome disapproval”). This paper explores a third possible explanation: public disapproval of the processes by which legislators decide how to vote on proposed legislation (“process disapproval”).A survey of a nationally representative sample of American adults revealed that Americans want representatives to make voting decisions to pay the most attention to the preferences of the general public and of the “issue public” (i.e., the people who feel strongly about the issue). In contrast, Americans perceive their representatives to pay the most attention to the preferences of their supporters, of their campaign donors,and of economic elites. Remarkably, Democratic and Republican citizens share the same view of how this aspect of government should and does function. The divergence between what the public wants and what people think the Congress does explains more of the public’s disapproval of Congress than does dissatisfaction with volume or outcomes. And the survey data suggest that disapproval of Congressional decision-making process may also shape attitudes toward the government as a whole and toward democracy as a system of government more broadly. Thus, the public’s perception of the legitimacy of a government may hinge importantly on how politicians explain their decisions.