Past Events

Gary King - Albert J. Weatherhead III University Professor - Director, IQSS - Harvard University

We show how to use matching in causal inference to ameliorate model dependence -- where small, indefensible changes in model specification have large impacts on our conclusions. We introduce matching methods that are simpler, more powerful, and easier to understand than existing approaches. We also show that propensity score matching, an enormously popular method, often accomplishes the opposite of its intended goal -- increasing imbalance, inefficiency, model dependence, and bias -- and should not be used in applications.  See http://bit.ly/causeI for papers and easy-to-use software to implement all methods discussed.

Location: Filene Auditorium, Moore Hall

Harry Enten '11, FiveThirtyEight, is a senior political analyst and writer for FiveThirtyEight, a data journalism website. He studies polling and demographic trends to try and tell readers who and why candidates and parties win and lose elections. Previously, he was a writer with The Guardian in New York. Harry graduated from Dartmouth College, where he was better known for providing snowfall forecasts to students via blitz. He still loves snow.

This public talk was jointly sponsored by the Dartmouth Institute for Health Policy & Clinical Practice, Government Department, Program in Quantitative Social Science, The Nelson A. Rockefeller Center for Public Policy and the Social Sciences.

For post-event coverage, see the article in The Dartmouth.

Location: Rockefeller Center 1930s Room

Seminar with Trevon Logan, Chairperson of the Department of Economics at The Ohio State University, and a Research Associate at the National Bureau of Economic Research  (joint with Lisa D. Cook and John Parman)

Using a newly developed measure of segregation, we provide the first estimates of the relationship between racial residential segregation and lynching in the post-Civil War U.S. We find that racially segregated counties in the southern United States were much more likely to experience lynchings of African Americans during the 1882 to 1935 period.

Location: Silsby 215

Research design seminar with David Cottrell, Postdoctoral Fellow in the Program in Quantitative Social Science, and Michael Herron, Professor of Government

Does waiting in line to vote in the current election reduce one's propensity to vote in future elections?  To answer this question, we analyze a unique voter file produced by EViD voting machines during the early voting period for Florida's 2012 general election.  The advantage of this voter file is that it not only identifies the individuals who voted during this period, but also identifies the precise time at which they cast their vote.  We leverage this information to determine if those who waited to vote after polls closed were less likely to vote than those who did not have to wait.

Location:  Silsby 215

Research design seminar with DJ Flynn, Post-Doctoral Fellow in the Program in Quantitative Social Science, Dartmouth College.

This project uses surveys and experiments to examine the causes and consequences of exaggerated beliefs about the power of the American president.

Location: Silsby 215

Seminar with Yusaku Horiuchi, Professor of Government, and Alexandra Woodruff '17.   

There have been numerous studies on a so-called “ballot order effect” – the tendency of candidates with favorable positions on a ballot to receive more votes. Given this concern, some governments (e.g., Australia) completely randomize the order of candidates' ballot positions. Yet, as far as we know, there has been no existing study on "side-effects" of randomization. We argue that the randomization of ballot positions increases invalid (spoilt/informal) votes, because not seeing one's favorable candidate at the top of ballot may discourage a voter to complete the ballot paper (or may invite an unintentional error due to additional cognitive strain). We use more than 60,000 polling-place level data from Australia's Lower House elections (1996-2013) and show that the percentage of informal votes significantly decreases when the candidate at the top of the ballot paper is the candidate who received the largest number of first-preference votes (i.e., the strongest candidate, but not necessary the winner under Australia' alternative-voting system). We argue that the randomization of ballot positions, which is aimed to adhere to the "fairness" principle of democracy, inadvertently undermines the "broad participation" principle of democracy. 

Location: The Drake Room at the Hanover Inn

The increasing centrality of the Internet and other technologies in the daily lives of people globally has generated a massive windfall of potential new data sources for social scientists. This one day workshop, hosted by the Neukom Institute, the Department of Sociology, and the Program in Quantitative Social Science at Dartmouth College, will advance the goal of developing new and innovative data sources and analytic strategies for the social sciences. The workshop will cover a broad range of substantive topics, from politics to gender, and will employ a wide range of methods including social media analysis, the use of UAVs, and morphometric analysis. In bringing together a diverse group of scholars committed developing new data sources, this workshop will advance the goal of fostering interdisciplinary connections that will carry forth outside of Dartmouth after the conclusion of the workshop.

See the complete program here.

Location: Silsby 215

Seminar with Jake Bowers, Political Science and Statistics and National Center for Supercomputing Applications at University of Illinois at Urbana-Champaign.

Background information can increase the precision of statistical inferences in experiments if we know the model relating the background covariates to outcomes. Machine learning techniques are built to enable and improve prediction given many variables. This paper proposes one way to use the power of machine learning to make experimental inferences more precise without any data snooping. 

Location: Silsby 119

Seminar with Alex Garlick from the Political Science Department at the University of Pennsylvania.

Description Lobbying is usually thought to be a moderating influence on legislator behavior. However, this paper uses bill-level lobbying data from two states in 2011-2014 to show that lobbying can lead legislators to vote on party lines. It highlights the polarizing effects of lobbying by policy demanding interest groups and increasingly politicized business interests.

Location: Silsby 119

Seminar with Jason Barabas, from the Department of Political Science at Stony Brook University. Many experiments use “opt-in” convenience samples from unknown populations, but does this mean that the results do not generalize? We conduct experiments that select and assign subjects to modes randomly in an attempt to (1) document patterns of nonrandom participation in experiments and (2) statistically adjust our estimates to offset selection biases.

Location:  Silsby 119

Seminar with Michael Link, Ph.D., SVP Measurement Science Institute, Nielsen & President, American Association for Public Opinion Research (AAPOR). Are we in the midst of a paradigm shift in the way we conceptualize, collect and analyze data on opinions, attitudes and behaviors? As surveys and polls face increasing problems, researchers are exploring new alternatives leveraging social media and other forms of "Big Data" to gain insights into attitudes and behaviors. The seminar will explore opportunities and challenges along this road and explore where it may be leading.

Location: Silsby 119

A seminar with Brendan Nyhan, Assistant Professor, Government Department, Dartmouth College. We show risks of post-treatment bias in turnout studies using voter files conditioning on registration and offer a new analysis approach to diagnose effects. The associated paper can be seen here:  http://www.dartmouth.edu/~nyhan/diff-reg-bias.pdf.

Location:  Silsby 215

Seminar with Michael Herron, Professor of Government and Chair, Program in Quantitative Social Science, and David Cottrell, Post-Doctoral Fellow, Program in Quantitative Social Science.

Location:  Silsby 119

Seminar with David Leblang who is Chair of the Department of Politics and J. Wilson Newman Professor of Governance at the Miller Center of Public Affairs and Professor of Politics at the University of Virginia.

We examine the factors leading to Germany’s decision to bail out Greece in 2012 and conclude it was political rather than economic factors that drove the decision making.  Specifically, we argue that the bailout was an attempt to protect German labor markets from foreign competition.  Read the complete paper by clicking here.

Location:  Class of 1930 Room, Rockefeller Center

Seminar with Jacob Montgomery Ph.D., from the Political Science Department, Washington University, St. Louis

Recent scholarship in political science has expanded our understanding of how personality affects political behaviors, attitudes, and learning. However, a major obstacle to expanding this research agenda is that many established personality inventories contain far too many questions for inclusion on surveys. In response, researchers typically select a subset of items to administer, a practice that can dramatically lower measurement precision and accuracy. In this paper, we outline an alternative method – adaptive personality inventories (APIs) – for including large personality batteries on surveys while minimizing the number of questions each respondent must answer.

Location:  Silsby 215

Seminar with Deborah Beim, an Assistant Professor of Political Science at Yale University, who received her PhD from Princeton University.

Conflicts between the Courts of Appeals are of central importance to the American judiciary. When circuits split, federal law is applied differently in different parts of the country. It has long been known that the existence of a circuit split is the best predictor of Supreme Court review, but data availability has constrained understanding of circuit splits to this fact. In this paper, we explore the ``lifecycle'' of an intercircuit split. We analyze an original dataset that comprises the universe of conflicts between Courts of Appeals that existed between 2005 and 2013, which includes both conflicts the Supreme Court resolved and conflicts it has not yet resolved. We show how long a conflict exists before it is resolved and how many go unresolved altogether, which conflicts are resolved soonest, and how a conflict grows across circuits.

Location:  Silsby 215

Seminar with Kentaro Fukumoto, who teaches in the Department of Political Science at Washington University, St. Louis.

If a unit is compromised (e.g. missing, non-compliance) in experimental data with blocking, some scholars recommend deleting the other units in the same block, saying this leads to unbiased estimates. Others, however, criticize the blockwise deletion, arguing that it can result in biased estimates if the compromising mechanism is not independent of the potential outcomes. This paper arbitrates this controversy by exact calculation of the simplest case and simulation.

Location: Silsby 119

Seminar with Allen Riddell, a William H. Neukom 1964 Fellow at Dartmouth College, who received his PhD in the Program in Literature at Duke University. This presentation explores how probabilistic models of text can be used to address longstanding questions about the relationship between the circuit courts and the Supreme Court.

Location: Silsby 119

Seminar with Eitan Hersh, an Assistant Professor of Political Science at Yale University, who received his PhD from Harvard. This presentation will show that the strategies political campaigns use to interact with voters are a result of the policy environment in which campaigns operate. Public policies about the collection and distribution of personal data affect campaign strategy and voter behavior.

Location: Silsby 113

First-Year-Orientation Open House: Mathematics and Social Sciences is an interdisciplinary program that focuses on the use of statistics and computing in the social sciences. Students who want to combine interests in a social science field with courses that emphasize technical skills — with particular emphasis on data analysis and computation — should stop by to learn about this program and its plans for the future.

Location: Leverone Field House

Orientation Information Expo: First-year students are invited to bring their parents and families to the very popular Expo where you can explore many different aspects of the College by stopping at the various tables and speaking with the representatives there. Do come by and learn more about the Mathematics and Social Sciences Program, and all that it can offer you in your studies!