Start Here
How to find a dataset for your research question
Once you have a question in mind and a theory you want to test, the next step is to find data that let you check the empirical implications of your argument. There are three ways to get there.
1. Use an existing dataset
Most of the time, someone has already collected data on the concept you care about. Find Data lists many of the most widely used sources in American politics, comparative politics, and international relations.
- Search by concept, not just by dataset name. Typing corruption or turnout shows every source with a relevant variable.
- Think about what one row is. A survey has one row per person; a cross-national dataset usually has one row per country per year (a country-year); event data has one row per protest, attack, or battle. Your research question determines which you need. Use the “What is one row?” filter.
- Look for replication data. If your question builds on a specific published article, the authors’ data and code are often public. Search the Harvard Dataverse for the article title or author, or check the journal’s website and the authors’ own pages. This is especially likely for recent articles.
- Not sure where to begin? The QoG Data Finder lets you search by concept across hundreds of datasets, and the QoG Standard Dataset merges 2,000+ country-year variables into a single file. For U.S. states, the Correlates of State Policy dataset plays the same role.
2. Collect your own data
Collecting your own data can be time-consuming, but it is also rewarding, and tools often exist to help automate the process. Know what the end product should look like before you start. For most projects it is a spreadsheet saved as a .csv file:
- the first row holds the variable names (one per column);
- each following row is one observation (one country-year, one person, one bill…);
- each column is one variable.
A file in that shape loads directly into R, Stata, or any other statistics program.
3. Combine datasets
You will often need to merge information from one dataset into another, for example adding GDP from the World Development Indicators to democracy scores from V-Dem. You may also need to aggregate information up to a higher level of analysis, such as averaging individual survey responses to the country level. To merge, both files need a shared identifier (country name or code, plus year). Watch out for countries that are spelled differently across sources.
Getting help
OU Libraries’ Research Data Support can help you locate data, and the library’s A–Z database list gives access to subscription sources marked “Through OU Libraries” on this site.