The Composition of Descriptive Representation

John Gerring | Connor Jerzak | Erzen Öncel

Paper Code Data

Who rules, and who is ruled? This is an essential question in the study of politics. In this paper, we bring new theory and data to bear on the question. First, we extend the Governance and Leadership Project (GLP) with a new round of coding: we characterize the gender, ethnicity, religion, and linguistic group of over 50,000 political elites around the world. We also gather data on the distribution of gender, ethnicities, religions, and linguistic groups for over 125 countries, allowing us to compare the distribution of group features in the population to the political bodies meant to represent them.

After creating this expansive data asset, we probe the reasons why some groups are better represented in political bodies than others. We approximate the complex dynamics by which people are selected into political bodies with a random sampling model.

Under this random sampling model, we derive the expected degree of representation that we would expect as a function of the size of the political body and the distribution of groups. Using this model, we find that every country in the world is, on aggregate, less representative than one would expect given the random sampling model, even though the random sampling model explains more than 50% of the variability in observed representation outcomes.

For a project like this, we cannot perform an experiment sending some (but not other) people randomly into political office. Therefore, we perform instrumental variables analyses in an effort to address concerns around confounding. We also perform extensive robustness tests to explore the sensitivity of our results to factors such as data missingness and hierarchically nested source of uncertainty.

We make accessible an open-source software package (available at this link), allowing others to compute theoretical quantities of interest introduced in the paper, such as the expected degree of representation and also the amount of representation not captured by our simple but powerful model.

References

John Gerring, Connor T. Jerzak, Erzen Öncel. The Composition of Descriptive Representation. American Political Science Review: 1-18, 2023.
@article{gerring2023composition,
  title={The Composition of Descriptive Representation},
  author={Gerring, John and Connor T. Jerzak and Erzen Öncel},
  journal={American Political Science Review},
  year={2023},
  volume={},
  number={},
  pages={1-18}
}
[Data][Code]

Back to Research
Back to Home