More gruel
There Are Consequences

There Are Consequences

The states that voted most heavily for Trump in 2016 avoided the first wave of the coronavirus outbreak. The virus swept through major metropolitan areas, in particular in the North-East where the highest density of people lives in the country. Even without a statistical analysis, you can see that a map of the number of coronavirus cases matches to a map of population density.

Coronavirus in the U.S.: Latest Map and Case Count - The New York ...
Coronavirus Confirmed Cases as of May 28, 2020. (Source: New York Times, retrieved May 28, 2020)
Population Density of the United States, 2011. (Source: United States Census Bureau, retrieved May 28, 2020).

This is to be expected and is mostly in line with the incidence seen in other countries – in Italy it was the densely populated North that was most heavily hit. In England, the South-East, including London, bore the early brunt. Thus the “Red Counties” dodged the first wave for the most part.

County level results of the 2016 US Presidential Election. (Source: Brilliant Maps, retrieved May 28, 2020)

The latest analysis of growth rates should put this assumption to the sword, however. The correlation between how heavily a state voted for Trump in 2016, and how fast the case rate is growing has shown a large rise over the last six weeks or so. Growth in cases is a measure of how well a disease is being controlled, and “Trump Country” is not controlling this virus well. The correlation between voting for Trump and growth is becoming clearer and clearer in the statistics, even if it isn’t being acknowledged on the ground.

The trend line (dotted) shows that the increase in correlation is 0.4% per day. That means that the current growth in correlation is over 10% per month. At current rates, the correlation will exceed 50% in September or October – not a good time for the impact on “Trump Country” on Trump’s chances in the 2020 election.

It isn’t just confirmed positive test rates that are growing, the trend can be seen in the growth in death rates as well.

With death rates, the growth is even faster, growing at 0.8% per day, or over 20% per month. This level of growth cannot continue indefinitely, as it would exceed 100% well before the election, however if the numbers are anything above 50% in October 2020 the consequences for Trump would be dire – people notice when a lot of people in their local communities are dying of a disease.


  1. Off topic,
    My Daughter was in the Twin Cities last Saturday…..then comes Monday evening. She was at home with us in St Peter MN on Monday afternoon when this all unfolded. She was supposed to visit friends this weekend in the cities. Not going to happen. I remember the protest of the Vietnam War in the seventies , saw them first hand in down town Houston. I don’t know what this is , outside agitators coming in starting shit.

  2. I am not going to comment on the veracity of this analysis. I have taken many stats classes, but I am no where near qualified to discuss this.

    What I can say is that this kind of analysis will soon be, if not already is, utterly useless, as more and more red states lie about the true numbers, or simply stop reporting them.

    The u.s. is a failed state. The tyrant and his cabal are by any measure the worst thing to happen to the country, and the planet, since WWII. They learned from any number of past dictatorships the value of controlling data. And they are getting real-time reinforcement of that value from china, russia, and brazil, to name a few.

    I am really wondering what historians will write, assuming the madman does not push the button, about how quickly the u.s. burned.

    1. I think we are in a crazy blip with Trump. If Trump loses in November as I expect, I think he and everybody else is going to be surprised how quickly Trumpism is erased from the Republican Party. This level of stupidity is self-defeating. We saw McCarthy disappear, then Goldwater, and soon Trump. The Democrats will have a couple of terms to make things sensible again and then we’ll get the next wave of Republicanism – let’s hope it is more the Country Club/Chamber of Commerce Republicanism than the yahoo crowd.

      1. I’d be nice if you are correct. However, IMO much longer than two terms will be required for a civil Republican Party to appear. The Republicans have gone so far to the nut-job right wing, that several Democratic terms will be required. After the debacle the Republicans created with the Great Depression, sensible Republicanism did not reappear until the Eisenhower Administration, a full twenty years, 5 presidential cycles. During the thirties the party flirted with fascism, American Firstism, and then Mccarthyism in the late 40’s and early 50’s. There was even an assassination attempt against FDR and an attempt at a coup. These were not directly associated with the R Party, but were elements of the “yahoo” crowd, that dominated the Party at the time.

        On the other hand, I do see a faint silver lining in that if there is a true blue tsunami in November, all the Republicans in DC will be so discredited that the Republicans will no longer be able to block and obstruct legislation as they were able to do so in 1993 & 94 and 2009 & 2010, so perhaps the Biden Administration (if it happens) will be able to accomplish some things. Even McConnell will hopefully be forced into retirement. This might shorten the rebuilding process.

        If Trump is not thoroughly defeated, a true miracle can only save the US.

      2. Neil, I don’t know what planet you are on, but the violence witnessed in the past few days is nothing compared to what is coming.

        I have posted before 8 possible outcomes in November and beyond, or at least leading up to the hypothetical election date. All but one lead to disaster, and that one is the scenario where the tyrant gracefully and humbly accepts a defeat at the polls, and then acts like a lame duck president until Jan 20th. We all know the chances of that happening.

        If he really believes he will lose an election, and said election can’t be rigged (it already has, BTW), releasing his cult members to protect him at all costs is automatic. If he “wins”, the only option is civil war to reverse the descent into fascism.

        There is no way, no how, this is going to have a happy ending.

  3. In Washington State, we have intrastate urban – rural contrasts that are similar to that portrayed above. Those contrasts are definitely showing up in the COVID-19 data for the state. The major population is located in the Seattle Metropolitan Area that largely is located in three counties, King, Snohomish and Pierce. The metropolitan area has approximately 2/3 of the state’s population. The rest of the state is much less densely populated. The remaining portion West of the Cascades Crest is rural to exurban, with several smaller urban areas and the Eastern portion is really part of the far west, is agricultural and is more like Idaho, Montana and Eastern Oregon. It is sparsely populated.

    One thing to note is that WA has large counties, unlike some other states. Some of our counties are larger in geographical size than many of the Eastern States. Even the three large population counties have significant rural areas. I use the county data because that is how the data is reported; more granular data is not readily available.

    The interesting thing is that in the three large metropolitan counties, the COVID-19 epidemic peaked and then after initially decreasing plateaued about three weeks ago. This last week the new case curve began slowly decreasing and the number of daily deaths is also slowly decreasing. All of the three counties are somewhat above the threshold to begin Phase 2 of Governor Inslee’s phased reopening plan. They continue to be under strong recommendations for “stay home, stay healthy.”

    On the other hand the worst COVID-19 hotspot is Yakima County, which is a rural agricultural county, with some meat processing and other agricultural plants. It is also one of the most conservative counties in the state. Its daily new incidence curve is far above those of the three Seattle Metropolitan Area Counties. It also is more likely to reject the social distancing guidelines. It does have a significant population center, the City of Yakima. The same is true for a few of the other agricultural counties. These counties have moderate population densities, relatively speaking.

    On the other hand, more than half of the counties are far more sparsely populated and to a large extent can be considered part of the ‘big empty’. These counties have progressed to Phase 2. We have one rural very sparsely populated county that has had no COVID-19 cases.

    Washington has much of the population diversity discussed in the above post and our situation shows the trends that Neil M. discussed. Fortunately, we have a good Governor, who has been forthright and honest, so we have had little of the political difficulties, that have been seen elsewhere. To be sure, there is some political opposition, but it is muted and the pressure has been released by the phased reopening plans.

  4. Interesting theory, Neil. I have zero background in statistical analysis so can’t offer anything except this one observation:
    The red state Data is misleading for many reasons. Late start in testing, selective data reporting, and misleading data compilation.

    Density aside (highly relevant), attitudes aside (refusal to mask, distance, and avoid groups), the refusal to accept the science behind viral spread is catching up with rural areas. Unfortunately, rural areas are less prepared to manage spikes in viral acuity. Suburban areas are better positioned for medical access but depending upon the partisan make up, may similarly have to confront consequences for Early reopening.

    Case in point.

  5. I like this bog because it attracts intelligent educated people. In process control you constantly run analysises and change parameters to reach your desired values according to the test data. A control loop.

    I hope rural people get smarter and change the data in a positive way. Not all of us are asses. Even in Ruby red Trump country a significant number of people have not drank the koolaidade.

    I think Democrsts can win rural voters. Something Stacy Abrams and Dough Jones proved. A true conservative view is desirable in governance.

    We do not fascism or racism. Which sadly the GOP has drifted into.

  6. (Sorry if this is a double comment, I may have submitted it somehow via a different method)

    I love your commentary, I find it frequently insightful and always interesting.

    But what is the source of the last two charts? What does “Trump 2016 correlation” mean as the solid line? I think it may be the daily percent increase/decrease in positive tests and or deaths in some segment of counties/states that voted for Trump above some threshold, but if so, what threshold? How was that threshold chosen? Also, linear regression can mislead as much as illuminate when used on time-series data like this; it should be analyzed using appropriate time-series analysis that takes account of the fact the data are auto-correlated. Further, in the absence of a similar analysis of the data from non-Trump 2016 voting areas there is no way to interpret these data. Finally, I don’t think any analysis of this type can be trusted unless it takes into account underlying differences in population and the problems with the data itself. For example; if “Trump Country” skews older a higher rate of death might be expected even with a lower rate of infection. Or the increase in positive tests could simply be a result of more testing in “Trump Country” as those communities are finally recognizing the urgency of the situation.

    I’m not saying there is no correlation, don’t get me wrong; I have no idea. Its certainly a hypothesis worth investigating. But these two charts, whatever their source, don’t make the case in any way.

    1. Hi Hans. The source of the last two charts is form data collected from “Worldometer” (

      The correlations were done using the “CORREL” Excel function, where the seven day growth rate for each state was correlated with the percentage of voters who voted for Trump in 2016.

      For example, Washington DC voted 4% for Trump, but had early growth that was larger than e.g. Idaho which voted 59% for Trump. Thus there was a negative correlation in the early stages of the pandemic, however the growth numbers today show a positive 29.8% correlation.

      The trend line is the trend for the change in correlation – the correlation is growing at roughly 0.39%/day currently.

      I’ve been watching this for two months – the early numbers were based on such small absolute numbers that the volatility was too high to feel any confidence in conclusions, but the numbers, with over 1.5 million positive tests and 100,000 unfortunate deaths now are starting to settle down.

      I am happy to send you the raw data if you want to analyze them by yourself.

      Neil M.

      1. Hi Neil: I don’t need the raw data, but I appreciate the transparency embodied in the offer. Given your clarification I now better understand what the chart is depicting, which is helpful. By “seven day growth rate” I’m thinking you mean the percent increase or decrease in the total number of cases from week to week?

        I caution you in putting much weight on the analysis you have done. There are all kinds of methodological reasons why a linear regression analysis of these data could be just as much misleading as useful. For example, there is good reason to assume a non-linear relationship because Pearson correlation is a bounded quantity (It cannot be outside of -1 and +1). In percentage terms, it cannot be greater than 100% (absolute correlation). Also, while there is no absolutely agreed upon intepretation of what constitutes low/medium/high correlation using the Pearson Correlation coefficient, I think most folks would agree that values of 0.3 or less represent relatively low correlation. So both of these charts could just as easily be interpreted as showing no meaningful correlation at all between the percent of voters who voted for Trump and the seven day growth rate.

        Again, I”m not saying there is no relationship. I’m just cautioning you that your analysis does not necessarily provide evidence for that relationship.

        In case you are wondering why it matters to me and why you should care what I think the obvious answer is I am just a dude on the internet, you probably shouldn’t! 🙂 But I am a master’s trained epidemiologist with many years of data analysis and health research methods experience I do think I have an “earned opinion” on the subject. Its just that, though, an opinion.

        Again, I appreciate your writing and look forward to future posts.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.