Abstract: Do empirical estimates in economics reflect the political orientation of economists? We show that policy-relevant parameters are correlated with economist partisanship as predicted from the text of published academic papers. Specifically, we build a model to predict the observed political behavior of a subset of economists which we then use to predict partisanship for all economists. Using our out-of-sample predictions, we show considerable sorting of economists into fields of research, and yet can detect differences in partisanship among economists even within a field, even across those estimating the same theoretical parameter. Using policy-relevant parameters collected from previous meta-analyses we then show that imputed partisanship is correlated with estimated parameters, such that the implied policy prescription is consistent with partisan leaning.

µþ¾±´Ç:ÌýZubin Jelveh is a Research Director at Crime Lab New York, a University of Chicago-affiliated research institute that partners with civic and community leaders in New York City and New Jersey to design, test, and scale promising programs and policies to reduce crime and violence. His research interests include the development and evaluation of prediction models intended to improve outcomes in the areas of domestic violence, gun violence, and child safety. He also studies the science production function, specifically the incentives that drive how research is presented.

Zubin Jelveh holds a BA in economics from the University of Chicago, an MA in quantitative methods in the social sciences from Columbia University, and a PhD in Computer Science from New York University. In a previous life, Zubin was a journalist covering economics for outlets like The New York Times, Condé NastÌý±Ê´Ç°ù³Ù´Ú´Ç±ô¾±´Ç, and The New Republic.

Digging Deeper: Representations for Fine-Grained Affective Text Analysis

How do we identify what kinds of sentiment and emotion a text evokes? While there is a long history of research on sentiment analysis, this talk describes new methods that draw on representation learning and deep learning to provide a more detailed understanding of a text and its affective associations. This encompasses methods that predict the specific emotions associated with a text, considering not only the semantic content but also the way the text is presented. For example, certain fonts and colors are perceived as more exciting, while others are more likely to convey trustworthiness. This also includes methods that better account for the sentence context in which words occur. The talk will conclude with an overview of applications and specific analyses resulting from this work.

Biography:
Gerard de Melo is an Assistant Professor at Rutgers University, where he serves as the Director of the Deep Data Lab. He has published over 100 papers on natural language processing and AI, and received Best Paper/Demo awards at WWW 2011, CIKM 2010, ICGL 2008, and the NAACL 2015 Workshop on Vector Space Modeling. Prior to joining Rutgers, he was a faculty member at Tsinghua University and a Post-Doctoral Research Scholar at ICSI/UC Berkeley. He received his doctoral degree at the Max Planck Institute for Informatics. Notable research projects include Lexvo.org, FrameBase.org, the Universal WordNet, and the Etymological WordNet. For more information, please consult .