Media coverage of foreign and minority groups

Do the media cover different groups of people differently? When they write about minority groups, does the average tone (positive, negative, etc.) of articles change? Do they frame similar issues in different ways depending on whether the group involved is domestic or foreign? Do any of these patterns change over time?

Media coverage has an enormous influence on how people think about different political issues. As such, learning about patterns and trends in this coverage is not just of general interest, but also has considerable policy relevance. Our research on this project follows three tracks:

For a look at similar work in this vein, see:


Ideology across time & space

How polarized are politics today? Are left and right further apart than they were two or three decades ago? Is "the left" more or less to the left than it once was? Is the U.S. left more or less to the left than the UK left?

Most scholarly research in this area has focused on Congressional speeches and votes. But these can only tell us so much: Congressional speech is heavily focused on the legislative issue under consideration and strongly driven by inter-party positioning.

This research project looks, instead, at political speeches aimed at general audiences. We draw on recordings of speeches available at outlets such as Youtube. Using these, we aim to develop a way to position speeches systematically on a left-right scale that is constant over time and across countries.

The project is jointly run with prof. Jaime Settle's SNAPP lab. For a look at what others have done along these lines, see e.g.:


Geocoding foreign aid

When countries receive foreign aid, does most of it stay in the capital? Do regions affiliated with ruling political groups receive a disproportionate share? Does food aid go to those parts of a country where need is highest? Many questions about foreign aid require knowledge about the targeting of aid projects within recipient states.

The AidData project, jointly run by the College of William & Mary, Brigham Young University, and Development Gateway has made major strides in the geocoding of aid projects using trained human coders. Examples of the results are here and here. However, the sheer scope of the AidData database of projects makes it impossible to rely on manual coding for all of it.

The state of the art in automated geocoding has improved dramatically in recent years. However, aid data presents unique challenges: aid projects may target locations so small or obscure they cannot be found in online databases (gazetteers); they may target multiple locations; they may target locations whose names have changed over the years, etc.

The goal of this research project is to develop an automated geocoder whose reliability and accuracy matches that of a trained human coders. The research has two main components:

For recent work using geocoding in political economy (the first one using AidData's geocoding!), see:


Debating foreign aid

Why do countries give foreign aid? Thanks to the research and data collection efforts of AidData and other scholars and organizations, we know more and more about where aid goes, and for what types of projects. Yet the factors that determine choices for or against particular recipient states or aid projects remain less than fully understood.

While development agencies largely control individual project decision, national legislatures (the U.S. Congress, the British Parliament, etc.) generally have considerable influence over broad policy outlines. Our goal is to identify the arguments for or against aid that shape legislative outcomes, and to uncover patterns over time and across donor states in the salience of particular arguments. These patterns, in turn, can be used to improve our understanding of aid policy decision-making.

We are currently systematically collecting references to foreign aid in the official legislative records of several prominent donor countries, beginning with the United States and the United Kingdom. Once the data collection stage has been completed, we will use machine learning techniques to classify such references into a number of different argument categories.

For related work, see:


Studying the international relations discipline

What issues do international relations scholars like to study? How do they study them? How relevant is their research to the key issues facing policymakers today? And how do new ideas spread through the discipline and into policy, domestically and across borders?

Working together with the Teaching, Research, & International Policy (TRIP) project at the College of William & Mary and the Global Pathways project coordinated by Wiebke Wemheuer-Vogelaar at the Free University of Berlin, we analyze connections across the international relations and policy-making literatures.

Among others, we study patterns over time and across countries in the topics studied, theories used and works cited by international relations scholars in different countries. In doing so, we combine information generated by human and automated coding with data about the contents of articles (frequent key words, context, etc.).

For related and similar work in political science and other disciplines, see: