Created by Dan Goodspeed
using data from New York Times and
visualization help from Flourish.
Some have asked for ways to donate to help keep the charts updated and create new ones. It takes me about an hour a day to update the eight charts. I'm not expecting much, but if you'd like, I made a few options. I also made a monthly email newsletter with chart updates exclusively for supporters. Thanks! -Dan
The deaths-by-partisan chart was made after dozens, perhaps hundreds, of emails and online comments requesting one be made in response to the cases-by-partisan chart. I was hesitant to make this chart as the correlation between partisanship and deaths, unlike with cases, is indirect. The only correlation is... in general, the more cases you have, the more deaths you'll have. This chart generally shows that.
Once someone gets COVID, however, there are a lot of factors that play a role in whether or not they'll die from the disease:
The partisanship of the state in which they reside is pretty much a non-factor at that point. That's why deaths-by-partisanship is a weak correlation. It also feeds into people's misconception that death is the only bad result to come out someone catching the virus. That ignores the millions of Americans who have survived COVID, but with long-term or often permanent organ damage.
All that out of the way, the numbers above are the total normalized* deaths per million for each state since July. A '100' means .01% of the state's population has died of COVID since July 1. July 1 was chosen as a continuation of the partisan new cases since June. The time from symptom to death tends to be 2-8 weeks. That averages out to about five weeks, or July 6 since June 1. I just rounded it back to July 1. The charts use the same political affiliation data as the cases chart, the Cook Partisan Voting Index.
* "Normalization" (perhaps better called "smoothing") means the abnormalities in the data were evened out. For example, if there were 10 days in a row of a few cases/deaths a day and then one day of 1000... that looks awful and frenetic on a chart like this, even when framed in a per-week display. In reality, that 1000 is just a backlog catch-up, so I normalized it by spreading the thousand over previous dates for a more even / more realistic data. It works similarly when the total number of cases/deaths drops one day. Likely a correction from a previous report, I just subtracted the difference over previous dates to numbers that are probably closer to reality.