Local Government Diversity Dashboard

Data Sources

Q: Where did we get the list of local elected officeholders? 

A: We used our continuously updated, comprehensive list of local government elected officeholders serving all communities of 1,000 or more in the United States, from 2013 to 2024. Local elected officeholders include top elected officials (e.g., Mayor, County Executive) and governing board members (e.g., Council Member, County Legislator). This data was sourced from Power Almanac. 

 

Q: Where did we get the race/ethnicity estimates for the general population? 

A: We used the U.S. Census’s American Community Survey (ACS) rolling 5-year estimates for each year. For 2023 and 2024, we compared estimates for local elected officeholders to the 5-year estimates from 2022 as they were the most recent available. We specifically used the following categories: 

• Asian alone or in combination with one or more other races 

• Black or African American alone or in combination with one or more other races 

• Hispanic or Latino 

• Native Hawaiian or Pacific Islander alone or in combination with one or more other races 

 

Methodology

Q: How did we estimate the race/ethnicity of local elected officials? 

A: We used a combination of three algorithms: 

1. Deepface  

• Collected images through searches for first name, last name, and state of all local elected officeholders (images did not necessarily include the actual officeholder) 

• Used a combination of “MtCnn” and “Ssd” models for facial detection and alignment to increase accuracy of analysis 

• Analyzed images using the facial attribute analysis feature 

• Individuals were coded as Black/African American if Black was the most likely predicted race 

• Read about the development of this method here 

    2. Who Are You (wru)  

• Predicted race/ethnicity of officeholders using their surnames 

• Individuals were coded as Hispanic/Latino if this ethnicity had the highest probability 

• Read about the development of this method here 

    3. ​Namsor  

• Analyzed first and last names using the "Name Diaspora” feature 

• Individuals were coded as Asian and/or Pacific Islander if the probability was at least 0.4 that one of the top two most likely ethnicities fell into specific categories 

• Read more about this method here 

 

Limitations and considerations 

Q: What are some limitations to this data and methodology? 

A: The main limitations to this data and methodology are: 

• Names and images alone cannot definitively determine an individual’s race or ethnicity. Although our method was validated against self-reported data from our surveys as well as manual coding, our estimates should be interpreted cautiously. 

• We were not confident in our ability to accurately estimate the proportions of some important groups, such as Native Americans, and therefore did not include them in the analysis. 

• We acknowledge differences within groups (e.g., Black vs. African American) but did not break these out further due to data constraints. 

 

Q: How accurate are these estimates? 

A: Our methods were validated against self-reported data from our surveys and manual coding. However, these numbers should be interpreted as estimates only.  

 

Q: How do we account for multiracial officeholders in our estimates? 

A: Like the census categories to which we compared our estimates for local officeholders, our racial/ethnic categories are not mutually exclusive. Local elected officeholders may be counted in multiple racial/ethnic groups. 

 

Q: Why did we choose the methods we used? 

A: After comparing the results of different methods to our self-reported survey data and manual coding, we found that certain methods were more effective for predicting a particular race or ethnicity. The facial attribute analysis of Deepface proved most accurate for identifying Black individuals, while wru was most accurate for identifying Hispanic or Latino individuals. Namsor was chosen because it allowed us to differentiate between Asian and Pacific Islander individuals, unlike other methods. 

 

Q: Are there plans to expand the analysis? 

A: Yes! We plan to expand our analysis to include local government civil service leaders (e.g., Head of Planning/Zoning, Head of Economic Development). We also plan to analyze the race/ethnicity of local leaders in each state to see how local government representation varies geographically. Finally, we plan to expand to include gender analysis in addition to race/ethnicity. Interested in seeing the representation of another local government role? Contact us at info@civicpulse.org.