Following the news cycle, it seems Shelter in Place is driven politically, with conservatives itching to get out and liberals sheltering in place (I’m overgeneralizing). We also know that urban environments help spread the virus, so it would make sense that these people would support shelter in place. Finally, we also know that urban areas tend to be more liberal. So is Shelter in Place really a political issue, or are we seeing urban areas hit hard, and therefor people in those areas (who happen to be liberal), support sheltering further?
If you know me, you know I like to play around with data, but to all of you a warning that I am not a data scientist, and this is really just me playing around. Please share any and all thoughts and resources to further this thought if of interest.
First thing I did was pull together some data around Shelter-In-Place, for this I used the SIP index provided by Safegraph, followed by political ideology, and urbanization. For ease, I did this by state as opposed to county, which would likely show more granular results. What I have looks like this:
The first thing I looked at was deaths per 100k and plotted them against the SIP index:
Not surprising to see a clear trend where the more deaths you see, the more likely you are to stay at home. The correlation coefficient we’re seeing here is 0.645.
Then I looked at % of conservatives in a state vs SIP Index:
Again, not surprising to see, but it’s pretty clear that the less conservative a state is, the more likely they are to shelter in place. Correlation coefficient here is -0.804.
Taking a look next at liberal % vs SIP Index:
Again, not surprising to see that the more liberal states are sheltering more. Correlation coefficient here is 0.658.
Now, let’s look at urbanization rate vs SIP Index:
Again, not surprising to see that the higher percentage of people living in urban areas show a higher shelter-in-place index. Correlation coefficient here is 0.788.
If you want to see these combined, it looks like this:
It looks very similar to when you compare against deaths/100k:
Of course, that’s because we saw that SIP Index and Deaths/100k are correlated.
Here’s where I get a little fuzzy. Can I normalize the percentage of people who are conservative against the percentage of people who live in urban areas and compare agains the SIP index to see a correlation? A high correlation would suggest that SIP index is driven by conservative values (or media) beyond simply their exposure from living or not living in urbanized places.
For this, I multiplied the percentage of conservatives against the percentage of people living in urban areas and compared against the SIP index.
What you see here is the correlation disappearing, with a correlation coefficient of -0.076.
This shows that perhaps Shelter-In-Place isn’t based as much on political values as the media might suggest, but just as much a function of people’s exposure to the virus due to proximity to urban areas.