6/6
If you have questions or comments, please reach out to me, Mike, or Asya!

Check out the paper here: papers.ssrn.com/abstract=537....

5/
The broader takeaway: the success of democratic reforms depends on how they’re implemented.

Even well-designed efforts to improve representation can be blunted by those with the power to carry them out.

4/
Incumbent protection was most common when officials had both motive and opportunity:
– Smaller cities
– Whiter populations
– Lower turnout
– Competitive elections

The consequences were stark. Incumbent protection deterred challengers and reduced Latino electoral success.

3/
Using a redistricting simulation algorithm, we show that cities overwhelmingly drew lines that placed incumbents alone in their districts.

In 63% of cities, the enacted map ranked in the 99th percentile or higher for avoiding incumbent pairings.

2/
The California Voting Rights Act (2001) pushed hundreds of cities to switch from at-large to district elections to improve minority representation, especially for Latino communities.

But it left incumbent city councils in charge of approving the new district lines.

Please reach out with questions or feedback!

CongressRA is still under development. We attempt to recreate the measurement creation process of Binder (2009) as a preliminary example of a use case. We welcome feedback on other ways to demonstrate the advantages of LLM agents.

We presented this paper at last week's "AI and the Study of Political Institutions" at USC's Political Institutions and Political Economy (PIPE) Collaborative. In our paper, we introduce CongressRA, an LLM agent for studying the US Congress to showcase our framework.

Specifically, we argue and try to show that LLMs, when equipped with predefined functions and data retrieval capabilities, can help researchers with data collection, processing, and analysis.

This is a better link: doi.org/10.1038/s415...