If Democrats are more pro-science than Republicans, and if science helps control COVID-19, then it sounds logical that US states with a Democratic governor will handle COVID-19 better than those with a Republican governor.

Does this bear out in practice? If the answer is "no", what could explain the apparent contradiction?

If there's no clear relation between the affiliation of the governor & performance with respect to COVID-19, I'm also interested if there's a relation between the affiliation of the population (i.e. if the state has more Republican/Democratic voters) & performance.

Edit: by "better", I mean flattening the curve of cases. This is the one and only metric I care about.

  • 12
    I think this is interesting and I upvoted, but feel the need to post the necessary qualifier: correlation does not imply causation. – Jared Smith Jun 30 '20 at 10:48

The metric the question mentions is 'flattening the curve of cases'. There is some debate on how to measure this, but the specific metric I've decided to use is to fit a polynomial curve to the graph of cases over time, starting from when each state recorded their 100th case. I then take the second derivative of this curve, and evaluate it at the current date (June 29th), giving us the rate of change of the slope. This is our 'curve-flattening score'.

To give an example, let's look at the graph of cases over time for New York and Texas, shown below. Clearly, in raw numerical terms, New York has far more cases than Texas. One way to measure the handling of the pandemic would be to look at this number. On the other hand, if we look at the trajectory of the curve, New York's curve is flattening off, while Texas appears to be getting steeper. This latter metric is what I will be using in my analysis.

enter image description here

Using this metric, we get the following heatmap, with a higher number representing a worse outcome. We can compare this to the next map showing the party of the state governor. Note that Alaska & Hawaii are omitted in the first map, but are included in the full data below.

enter image description here

enter image description here

Looking at the governor of each state, we can see that the three states with the lowest score are are Texas, Florida & Arizona, all with Republican governors, although these are then followed by the Democrat-governed California and Illinois. Of the worst ten states, seven have Republican governors. Looking at the states judged to be handling the pandemic best by this metric, all of the top five, New York, Michigan, New Jersey, Louisiana, and Virginia are governed by Democrats, and eight of the top ten.

Removing outliers (values outside [-150, 150]), the average Democrat-governed state has a 'curve-flattening score' of -7.66, while the average Republican-governed state has a score of 20.05. This, again, suggests that Democratic governors are handling the pandemic better in their states than their Republican colleagues, based on this metric.

Based solely on this limited analysis, then, it seems that Democratic-governed states are flattening the curve of cases better than Republican-governed states.

Full data:

           State       Value       Party
        New York -270.716627    democrat
        Michigan -141.942838    democrat
      New Jersey  -82.264133    democrat
       Louisiana  -77.507513    democrat
        Virginia  -39.734256    democrat
        Maryland  -34.473446  republican
     Connecticut  -29.163433    democrat
    Pennsylvania  -21.205778    democrat
       Wisconsin  -19.795953    democrat
    South Dakota   -2.257446  republican
         Vermont   -1.847893  republican
      Washington   -1.718445    democrat
           Maine   -1.501176  republican
          Alaska   -0.931880  republican
    Rhode Island   -0.076976    democrat
   West Virginia    0.298186    democrat
   New Hampshire    0.380728  republican
          Hawaii    0.660018    democrat
  North Carolina    0.816185    democrat
         Montana    1.903081    democrat
    North Dakota    2.529137  republican
         Wyoming    3.486826  republican
        Kentucky    5.186839  republican
      New Mexico    6.325482    democrat
        Delaware   11.543138    democrat
        Colorado   12.881185    democrat
         Alabama   13.993028  republican
        Arkansas   14.869030  republican
        Missouri   14.881671  republican
          Oregon   16.153599    democrat
         Indiana   16.677088  republican
           Idaho   18.562333  republican
            Utah   19.872332  republican
       Minnesota   24.596237    democrat
     Mississippi   25.444759  republican
        Nebraska   26.300690  republican
          Nevada   28.483117    democrat
       Tennessee   37.187872  republican
        Oklahoma   40.680940  republican
            Ohio   41.395639  republican
            Iowa   44.830352  republican
          Kansas   48.995664    democrat
   Massachusetts   49.924742  republican
         Georgia   71.282481  republican
  South Carolina   74.612648  republican
        Illinois   99.889327    democrat
      California  187.667507    democrat
         Arizona  236.048719  republican
         Florida  356.647006  republican
           Texas  406.084270  republican
  • 11
    It could also suggest that Democratically-run states were particularly poor out of the gate, since the shape of the curve is also dependent on the initial performance. Lies and damned statistics, and all that, though especially damning is the fact that experts predicted the spike would happen in the states "opening up" more freely, and they were pretty much sneered at by the governors of those states. Note that when I say "it could," that doesn't mean I would necessarily stake out that position, myself. – PoloHoleSet Jun 29 '20 at 13:23
  • 11
    I am extremely dubious about fitting a polynomial to a curve that may be different shapes at different times due to policy changes. For example, the all-US curve appears to initially exponential, with the shape changing as lockdown orders were issued. – Patricia Shanahan Jun 29 '20 at 14:44
  • 3
    "New York's curve is flattening off, while Texas appears to be getting steeper. This latter metric is what I will be using." This is a garbage metric. The shape of Texas' curve is looks a lot like the shape New York's early curve. Likely it will flatten just like New York. – Paul Draper Jun 29 '20 at 23:20
  • 5
    @divibisan if every single man, woman, and child got COVID-19 in April, you'd give them that perfect score now. That should tell you that there is something seriously wrong with your metric. – Paul Draper Jun 30 '20 at 0:08
  • 7
    What mathematical or statistical justification do you have for fitting a polynomial curve or taking a second derivative as a metric? Isn't the traditional model the logistic curve? nctm.org/Classroom-Resources/Illuminations/Interactives/… – qwr Jun 30 '20 at 22:56

I calculated average case increases for the last 60 days, and took the ratio of that to all cases. Thus, the smaller the ratio, the more it indicates that the current case increases are far less than the peak, hence indicating a flattening of the slope.

The data is below.

enter image description here

I then ran a two-sample t-test for means. There is overwhelming evidence that the mean for Democrat-led states is smaller than the corresponding mean for Republican-led states. The means are (D: 0.009, R: 0.011) and the p-value under the null hypothesis of equal means is 0.005995, which is evidence against the null hypothesis.

One could argue that NYC, New Jersey, Hawaii, Guam, Virgin Islands, Northern Mariana Islands, and Puerto Rico are obvious outliers for different reasons (NYC/NJ being hit the hardest first, and the other being islands). I removed them both, re-did the analysis, and found that the p-value increases (obviously) but under the null-hypothesis of equal means, the p-value is 0.12. This is still small, and combined with the case where we did not remove outliers, speaks to some evidence that Democrat-led states have done better at flattening the slope.

Conclusion: Yes, the data does suggest that on average, Democrat-led states are doing better at handling the pandemic than Republican-led states, in terms of current performance versus peak performance.

  • 6
    Massachusetts, Maryland, Vermont, and Puerto Rico are Republican-controlled??!?! That's, ummm, simplistic. At best. Vermont, for example, voted 57-30 for Hillary Clinton in 2016, Maryland 60-33 for Clinton. Kansas, which you claim is Democrat-controlled, voted 56-35 for Trump. And by limiting your data to 60 days, you've removed the March COVID disaster in Democrat-controlled New York and New Jersey. That's a conspicuously interesting choice of dates... – Just Me Jul 1 '20 at 11:48
  • 5
    Massachusetts, Maryland, and Vermont are amongst the better performers of "Republican-led" states. Kansas is amongst the worse Democrat-led performers. I'd be happy to switch them as per your desires, but you do realize that will only strengthen the conclusion? Secondly, yes the limit of 60 days has indeed been chosen to remove the March cases of NYC and NJ.... that's .... kinda the whole point buddy. We want to measure ability to solve a problem, not the size of the problem. Read OP's question. Do you need help determining what "flattening the curve" means? I'd be happy to help, just ask. – Jaood Jul 1 '20 at 12:01
  • 4
    The point was you can't use statistics to "prove" a point when your selection criteria for inputs is downright sloppy. Discounting NY and NJ failures is deliberately cooking the books. Might as well ask how Brazil did in the last few minutes against Germany in the 2014 World Cup. Yeah, Brazil did outscore Germany 1-0 in the last few minutes of that match. Too bad the rest of the match also counts. – Just Me Jul 1 '20 at 15:03
  • 1
    @JustMe The word "prove" does not appear in my post, so why are you using quotation marks? What a pathetic and transparent attempt at a strawman-fallacy. Secondly, of course my selection criteria is sloppy, but it is a criteria, and using it, I did reach a conclusion. Probablistically, that is still evidence, since my criteria wasn't fueled by bias (as evidenced by the fact that your corrections to my criteria made my conclusion stronger!). Thirdly, I modified the analysis to include more of the data from March. Guess what? Same conclusion. NY/NJ perform worse but overall Dem-states do better. – Jaood Jul 1 '20 at 15:30
  • 1
    @JustMe Regarding the selection criteria, I am using party affiliation of the governor. This is not perfect, but the correlation is strong enough for it to not matter too much. But you are very welcome to provide me an alternative selection criteria, and I will gladly use it. I doubt you will though, because you know very well the conclusion will not change. – Jaood Jul 1 '20 at 15:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .