# Advice on normalizing census data in choropleth maps

I'm trying to figure out the best way to normalize census data in two choropleth maps showing the population of people with hearing loss. Their geometries:

1. The states of the United States.
2. The census tracts of one state.

For the first map (the states of the United States), I have the following data fields:

• Total population
• Total population with hearing loss

Ways I can think of to normalize the data:

• Divide the total population with hearing loss within a state by the state's total population
• Divide the total population with hearing loss within a state by the total population with hearing loss in the United States

For the second map (the census tracts of one state), I have similar data fields:

• Total population
• Total population with hearing loss

Ways I can think of to normalize the data:

• Divide the total population with hearing loss within a census tract by the census tract's total population
• Divide the total population with hearing loss within a census tract by the total population with hearing loss in the state

Essentially the same question for both maps. I see mixed advice when I Google this. What is the significance of each approach, and is there a clear better choice?

• It depends, what do you want to show with your map? Mar 27, 2020 at 16:02
• I want to show the amount of people who are deaf or hard of hearing. I am uncertain which normalizing approach I should use. Mar 27, 2020 at 20:10
• I think what inc42 was asking is "What do you want to say with your data?" There is no objectively wrong/right way to normalize--each strategy provides different information which will be useful (or not) toward answering different questions. If you just want to show amounts, don't normalize at all. It's also fine to show multiple normalization strategies, as each might reveal different nuances in the data. My advice would be to think about what your message(s) is, then use whichever strategy supports it best (while mentioning the caveats, of course).
– Jon
Mar 29, 2020 at 18:21
• Have you seen this discussion How to normalize choropleth maps of census information?? Additionally I may refer to these articles Normalization | Mapping Rates, Not Totals and Create and use a choropleth map. Mar 30, 2020 at 5:55

I think the first approach (divide the total population with hearing loss within a state by the state's total population) is the best way to go.

Here is my take on the different ways one could possibly display this data:

1. No normalization: This is non-ideal, because states with larger populations will naturally also have more people with hearing loss (and those with smaller populations fewer). The final product will likely look very similar to a simple map of population by state, and thus not be very informative.

2. Normalize by total state population: This accounts for problem mentioned above by accounting for the differences between states. Here you are essentially mapping the percent of a states population with hearing loss (aka per capita hearing loss by state). This map will be more informative than the first, because you can easily compare between states with vastly different populations.

3. Normalize by total US population with hearing loss: In my mind this is not really normalization. You are now showing what percent of all people with hearing loss (nationally) live in each state. If that's what you want to show it may be a good option. But you're essentially back to the same problem: States with larger pops will generally have a larger share of the national population with hearing loss, and the map will look very similar to a population map.

edit: some visuals