Covid in the states

 Andrew Gelman wrote about my post of December 1, 2021, which was about the state-level relationship between Joe Biden's vote share and Covid hospitalization rates.  I had suggested that there was a real but indirect relationship--Republican elites have been less supportive of vaccination than Democratic elites, so vaccination rates are lower in Republican states, so hospitalization rates are higher.  The comments on Andrew's post raised some points which I'll consider here.

1.  Does the relationship change with the seasons?  In hotter states, people tend to stay inside during the summer; in colder states, people tend to stay inside during the winter.  Since Democratic states tend to be colder, this suggests that they'd have lower Covid rates in the summer and higher in the winter.  My December post showed a weaker relationship than had existed in the past, so maybe that was a sign that things were changing?   Here are the correlations of hospitalization rates with Biden vote share at different dates:

March 7, 2021           .28    (.13) 

Late August 2021    -.42    (-.46)

Late Sept  2021       -.63   (-.70)

Late Oct  2021        -.51   (-.53)

Late Nov                 -.25   (-.23)

Feb 7, 2022             -.33   (-.57)   

The first column includes the fifty states plus DC, while the second (in parentheses) excludes DC.  The relationship has been negative at all times starting in August 2021.  But was it weaker in the winter months?   If you include DC, late November and early February are similar; if you don't, late November looks like an exception to the usual pattern. 

2.  Why is DC an outlier?  It's an outlier in terms of the vaccination/politics relationship:

 That's probably mostly because of racial composition:  About half of the DC population is black, higher than any other state.  Black voters are 90+% for the Democrats, but black people have lower vaccination rates than whites.

DC is also an outlier in terms of the vaccination/hospitalization relationship:

 

 It hasn't always been such an extreme outlier, but it does seem to be consistently above what would be expected.  Hospitalization rates are based on the number of people hospitalized in that state, not the usual state of residence of the patients.  So the DC numbers may be affected by the inclusion of people who live in the Maryland or Virginia suburbs (assuming there are more hospitals in the city than in the suburbs)

  3.  How much does DC affect the correlations?  I've already given the figures for the politics/hospitalization relationship.  The vaccination/hospitalization correlations:

                            included   (excluded)

March 7, 2021          -.37   (-.44) 

Late August 2021    -.59    (-.60)

Late Sept  2021       -.75   (-.76)

Late Oct  2021        -.48   (-.46)

Late Nov                 -.15   (-.13)

Feb 7, 2022             -.58   (-.63)  

Including or excluding DC doesn't make much difference.

The Biden vote/vaccination correlations:

                           included   (excluded)

March 7, 2021          -.03      (-.04) 

Late August 2021      .86      (.88)

Late Sept  2021         .86      (.89)

Late Oct  2021          .86       (.89)

Late Nov                   .84       (.87)

Feb 7, 2022               .83       (.86)  

Again, it doesn't make much difference.

4.  Were Republican politicians following rather than leading "the base"?   To some extent--in late 2020, polls already showed party differences in intent to get vaccinated.   But this wasn't an issue like abortion or gun control on which a lot of people have long-standing convictions that are hard to change--if Republican politicians had urged their supporters to get vaccinated, I think that would have made a difference.