Publication Type: Paper/Poster ProposalAbstract: Research shows that responses to survey questions can be influenced by whether respondents perceive themselves and their interviewers to be of the same race. Several studies have shown that African-American respondents who perceive they are being interviewed by white interviewers are likely to give substantively different responses from those of his/her counterparts who perceive the interviewer to be African-American. These findings are especially relevant when the content of questions directly relates to racial attitudes, perceptions of prejudice, and experiences related to race and ethnicity. With awareness of this potential interviewer effect on measurement for surveys of this kind, a university-based survey center developed an interviewer/respondent race-matching methodology for an annual survey of racial attitudes that began in 2003. While there are clear benefits to using this matching protocol, such as encouraging respondent honesty and frankness, there may be latent pitfalls for such designs. A survey design that incorporates interviewer/respondent race-matching protocol may interject unique problems into the processes of a call center. Thus to explore this issue further, we developed an exploratory qualitative inquiry to focus on over thirty interviewers' reactions to collecting data on race-related questions and especially their interpretations and sensitivity to the race-matching process. We were particularly interested in instances in which the race-matching design breaks down and respondents perceive interviewers to be of a different race. We argue that while survey researchers have made clear strides in understanding the effects of race-matching on measurement, there is much to learn about the potential for marginalization of interviewers. Moreover, given that effective interviewing is one key to achieving quality data in telephone survey operations, it is critical to understand the potential positive and deleterious outcomes that such survey procedures may have on the interviewing staff.
Publication Type: Conference Paper/Unpublished ManuscriptReview Method: Peer ReviewedAbstract: There is general agreement that college education confers political resources that reduce the costs and increase the benefits to participation. However, this consensus is not without challenge. Alternatively, individuals may attend college as a result of the same early life experiences that also influence their likelihood to participate in politics. The robust correlations between the two may in fact reveal a spurious causal relationship. Propensity score matching is increasingly being used in political science and elsewhere, to control for such selection confounders in the estimation of causal effects. In a pathbreaking piece, Kam and Palmer (2008) employ this technology to assess the effect of education on participation. In a major reversal, after p-score matching, the authors find no effect at all. We replicate Kam and Palmer's analysis to assess the robustness of their null finding. We use a cutting-edge technique, genetic matching, to show that the authors' zero estimate is highly sensitive to model specification and nearly-monotonic improvements in covariate matches. Generally, we show the fundamental difficulty of using p-score matching to estimate causal effects in observational data where selection bias is present.
Publication Type: Conference Paper/Unpublished ManuscriptAbstract: The college admissions problem studies many-to-one matching in bipartite graphs (each college is matched to the many students who will attend it). Traditional formulations of this problem assign a quota (the maximum enrollment allowed) to each college, and a great deal of research has focused on matchings which are stable in a game-theoretic sense with respect to the preferences of the colleges and the students. In this talk we will look at cases where the number of students enrolling at each college is desired to be equal (within 1 student) across colleges, without regard to preferences. This problem is connected to current undergraduate research in partially ordered sets with the normalized matching property.
Publication Type: Conference Paper/Unpublished ManuscriptAbstract: Since Granovetters frequently-cited work on the role of social networks in the labor market, many have tried to document the relationship between networks and labor market outcomes. However, the results of these studies are quite inconsistent. In this paper I specify two distinct theoretical mechanisms through which networks could affect labor market outcomes, and outline how each interacts with important structural features of networks and other labor market practices. Results from a computational experiment using these mechanisms accounts for many of these formerly inconsistent findings. Specifically, results confirm that context, in the sense of global network structure, strength of employment relationship, and employers preferences, has profound consequences for careers, the distribution of opportunity through the population, and the meaning of good and bad networks. These theoretical advances are particularly important in light of evidence that there has been a general decline of career jobs, a trend that puts renewed pressure on the job-matching process and those extra-organizational institutions--such as social networks--that play a role in job-seeking and hiring.
Publication Type: Conference Paper/Unpublished ManuscriptAbstract: This paper attempts to answer the question of whether ethnicity of the celebrity and match-up level between celebrity and product in a magazine ad by the advertiser has any effect on attitudes to the ad, attitudes to the brand, and purchase intention. An experiment was conducted where subjects were given four versions of ads - ads featuring Tiger Woods for sports drink, ads featuring K. J. Choi for sports drink, ads featuring Tiger Woods for ball-point pen, and ads featuring K. J. Choi for ball-point pen. The results of the study suggest that an ad with high relevance between celebrity and product can induce more positive Aad than an ad with low relevance. Results also show that Korean subjects were more likely to have more positive Aad, Ab and Pi for the advertisement featuring an American celebrity endorser.