AY 2018-2019 Compute Infrastructure 

MDI provides computing infrastructure to support public policy research. Click here for details.

Below are a few projects that have used the infrastructure. Contact mdiresearch@georgetown.edu if you have a project that needs compute infrastructure.

Parallel Processing with RStudio to Identify Potential Fraud in 2017 Honduran Presidential Election
Joel Simmons, Associate Professor in the Walsh School of Foreign Service and Irfan Nooruddin, Hamad bin Khalifa Professor of Indian Politics and Faculty Chair of the Walsh School of Foreign Service, worked with MDI to research the 2017 Honduran Presidential Election. Allegations exist that the 2017 Honduran presidential election was marred by massive fraud. Specifically, some observers allege that during a computer glitch that stalled ballot counting for thirty-six hours, the incumbent president Juan Hernandez manufactured votes to win re-election. Hernandez denies the accusation and claims that the glitch was a random, unexpected computer error. We distinguish between these two possibilities by estimating a finite mixture model using data from all 18,100 polling stations in the country.  They hypothesize that Hernandez's vote shares in the individual polling stations come from one of three component distributions: One with no fraud, one with incremental amounts of fraud, and one with extreme amounts of fraud. Our results show that (1) fraud almost certainly occurred, (2) the fraud was probably decisive - Dr. Simmons and Dr. Nooruddin estimate that 53,000 votes were stolen, compared to Hernandez's margin of victory of 52,000 votes, and (3) the probability a polling station belongs to the fraudulent components spikes dramatically immediately after the glitch.  The last point indicates that the "glitch" wasn't actually a glitch at all, but rather pretense for stealing the election.

Latin American Parliamentary Debates Database
Jennifer Tobin
, Associate Professor in the McCourt School of Public Policy, created a searchable database of parliamentary debates over time (for different periods) for Argentina, Belize, Chile, Costa Rica, Mexico, Peru, and Uruguay with the help of MDI. Some of the most widely studied questions in Political Science focus on what politicians say and how they say it. Scholars who study the United States and Great Britain have access to the daily records of the proceedings and debates in the US Congress and the UK Parliament. In developing countries, access to parliamentary proceedings is much more limited. In Latin America, many countries allow public access, but the ability to search the proceedings tends to be quite onerous. The Latin American Parliamentary Debates Database is a first step towards creating the equivalent, for Latin America, of the US Congressional Record and the UK’s Official Report of Parliamentary Debates. To begin with, we have scraped the available parliamentary debates for all Spanish- and English-speaking countries in Latin America.  

Using OCR to Process Legislative Documents
NaLette Brodnax, Assistant Professor in the McCourt School of Public Policy, is using Optical Character Recognition (OCR), a technology that enables you to convert and digitize different types of documents into editable and searchable data, to analyze the prevalence of education policies that the US state legislatures introduced following the Race to the Top education grant competition. From 2010 to 2012, the US Department of Education awarded $4 billion in competitive grants for state-level reforms meeting numerous criteria for innovation. Most analyses of policy diffusion often limit the utility for comparing a large number of cases where states adopt only part of another state’s policy, or the assessment of the magnitude of similarity between states. To address these limitations, she analyzes the extent to which these innovations appeared in state policies by constructing a corpus from a query of over 40,000 education bills introduced by state legislatures from 2009 to 2016. Dr. Brodnax finds that evaluation policies spread from states that were awarded competitive grants to those that did not, suggesting that states learn from, rather than compete with, states for the grant funding.