It is not surprising that moving out of lockdown resulted in more COVID-19 cases. In addition, the increasing number of tests undoubtedly increases the number of known infected.
But could there be something else going on?
Could increased use of air conditioning, particularly in the southern tier of states, be a significant driver of increasing number of COVID-19 cases?
This blog will attempt to help answer this question.
So where is the virus really spreading? A good way to see the problem locations is to view the percentage of positive tests. A worsening epidemic is signaled by a higher percentage of positives, assuming there is widespread testing. Positive percentage is far better than number of positive tests, which, of course, varies by the amount of testing.
Here is a plot of the positive percentage on July 7th. The big problem states were Arizona, followed Mississippi, Florida, South Carolina, Texas, Georgia, Alabama, Nevada, and finally Idaho.
Below is a different type of plot that shows the same thing, but provides the actual numerical values. The bottom line: the situation is far worse for states along the southern tier of the U.S. Arizona is the worst, with Mississippi and Florida right behind. These are states with very different demographics.
But what do these states have in common? Some media outlets are pushing the fact that most of these states are dominated by the Republican party and have been quicker to open up. But they have something else in common: these states have had high temperatures with a lot of air conditioning use. (And no, there is no reason to think that heat turns people into Republicans).
If we look at the high temperatures in June (shown below, NOAA division dataset), southern Arizona (including Tucson) is the nation's hot spot--and yes, it is the hot spot for COVID-19 as well. Mississippi, South Caroline, Florida, Texas are all very hot. And according to U.S. Census data nearly all homes and most restaurants in these states have AC.
And an independent graphic, showing the high temperatures averaged over the 30 days ending July 7th (Climate Prediction Center), has a similar pattern. Arizona has the highest temperatures.
So let us consider a hypothesis: the rapid warming in late spring led to greatly increased use of air conditioning in homes, stores and restaurants in the warm, southern tier states. More people are thrust into interior spaces with recycled, recirculating air that increases COVID-19 spread, something described in several research papers. And the cooler, drier conditions associated with air conditioned spaces are favorable for COVID-19, and the blowing air spreading COVID-19 containing droplets and aerosols.
Now is this hypothesis consistent with observations? We can begin by looking at the total tests and percent of positive tests in Arizona (see below). Tests went up substantially in May and June, but so did the percentage of positive tests, which has progressively risen since mid-May (the largest increase was in mid-June)
And this is exactly when temperatures surged to way above normal in southern Florida (see below). And Florida has terrible humidity as well. Folks were forced to flock to air conditioned spaces.
You want something more rigorous? No problem.
If I was writing a paper on this topic, I would present a scatter diagram plotting the temperatures against positive percentages of COVID-10. And I have done exactly that below. Specifically, I found the June average maximum temperature for every state in the continental U.S. and its corresponding positive percentage for COVID-19 (Y-axis percentage, X-axis is average high temperature). Each state is shown by a blue dot. I only plotted states with max temperatures in June of 75F or more, which excluded a handful of states that are very cool and have very few air conditioners (e.g., WA, OR, and Montana).
I also plotted a best-fit line (red). There DOES appear to be a relationship between COVID-19 infection rates and temperature. The correlation coefficient is .69, which suggests this relationship explains about 48% of the variability. That is quite a bit. The point in the upper right corner?--Arizona.
Now certainly there are a number of factors that help explain the variability of COVID-19 infection rates around the U.S. But I do think the above results are very, very suggestive that very warm temperatures result in increasing infection rate. Not because the virus likes warm temperatures (it does not, as shown by a number of studies), but because warm temperatures push people indoors into air conditioned spaces in which spread is greatly enhanced. Restaurants and bars are probably key here.
In warmer climates, summer is the time when folks huddle together in confined spaces and thus the greatest potential for COVID spread. The implication of all this is that the situation might be expected to worsen over the southern tier states and into the warm/humid areas of the southeast over the next month or so, but improve during the fall. Clearly, there is reason to avoid air conditioned interior spaces during a COVID epidemic, and dining should mainly limited to outdoor spaces, which should be quite safe.
The United States is not the only place on Earth with hot weather and air conditioning. If there is a need for a control group, there are plenty of examples to sway an argument either way. There are plenty of wealthy nations in Asia that swelter all year and have reliance on AC with low Covid cases (currently). There is also Brazil, which the majority of citizens do not have AC at all but is being crushed by the pandemic. So, pick whichever best proves your point....
ReplyDeleteBeing cooped up indoors with recycled air certainly is a factor, especially with buildings that have low levels of fresh air exchange due to efficiency. Its one of many environmental factors that contribute to a degree, but are certainly not unique to the USA.
Overall, the problems related to Covid-19 in the USA started with as well as remain with selfish attitudes and poor leadership.
I agree. And I cringe when I hear people say that there's more positives because there's more testing. The true number lies in the percentage of positives to the numbers tested. Those percentages are going up and those are the numbers we need to be looking at.
DeleteCompare it with countries with similar temperatures as south. e.g. in Asia for example.
ReplyDeleteCliff, this looks like something straight out of spuriouscorrelations.com. Rather convenient to omit Oregon and Washington, as they won't look pretty on your regression line. I don't think I would call your analysis "rigorous". Perhaps you should include other countries with high AC usage. Take Japan, for example. They have the highest per capita AC in the world and we don't see rates rising there. Let's not take the focus off of the true cause of spiking rates throughout the American south: opening unsafely with low mask use.
ReplyDeleteI don't think this is a spurious correlation. Adding Oregon and Washington does not change things much...but it does not make sense to include them....there is so little AC availability. Obviously, there are all kinds of cultural and other factors that come into play. And you can't wear masks while eating and drinking. My essential point is that inside restaurants/bars should not be open for high density usage when one can not open windows and flush out with outside air. You disagree?
DeleteI think Washington might be very interesting. From what I understand, most of the increase in positivity rate in WA is coming from the Eastern part of the state, where it is generally hotter, and where air conditioning is more prevalent. A county-level analysis that looks at temperature, positivity rate, and population density as a potential interaction or conditioning variable might be enlightening.
DeleteThanks for sharing Cliff. I was just having a conversation with a cousin in Phoenix and was hypothesizing that their summer is equivalent to our winter in terms of driving everyone indoors. Pair the drive indoors with a re-opening of services and we have a disaster. Makes me wonder whether the same will happen in northern regions when our weather gets colder and drives more people into recirculated heated indoor air.
ReplyDeleteYes, and a very compelling argument for public indoor mask use for everybody.
ReplyDeleteDr Scott Atlas suggests a correlation with COVID-19 infection in Northern Mexico and the Southern counties of California, Arizona, and Texas. I would expect Northern Mexico to have the same temperatures but do they have the same air conditioning availability/use?
ReplyDelete"The correlation coefficient is .481, which suggests this relationship explains about 23% of the variability." [sic]
ReplyDeleteThe scatter plot reports R^2 (coefficient of variation) = 0.481
The correlation coefficient (R) = 0.684
absolutely right...thanks, fixed. Need better reading glasses.
DeleteThanks for the analysis, Cliff. I've been wondering if extended dry, hot conditions might lead to an increase in a certain kind of aerosolized spread--that is, virus that initially is carried in droplets which then dry out and leave virus attached to dust particles that can be stirred up again into the air. I saw one article a while back that mentioned this kind of spread, but most current discussion of aerosols seems to focus on moisture and droplets.
ReplyDeleteI agree that low humidity (independent of temperature) is key to aerosolized virus transmission. Low humidity in heated indoor spaces during winter helps accelerate the seasonal flu. Meat and vegetable packing plants, notorious for COVID transmission, try to maintain low humidity to preserve their product and prevent food-borne illnesses.
DeleteCliff, I'm not epidemiologist, but it seems to me that a more interesting parameter to correlate with would be Rt ("average number of people who become infected by an infectious person"), not positivity. What if you took the Rt modeling from https://rt.live/ (also a good source for positivity modeling) and correlated it with state average maximum temperatures?
ReplyDeleteLooking at just the two states that you graphed separately, Arizona's modeled Rt peaked in late May and Florida's peaked at the beginning of June.
They don't use AC in AZ in homes, they use swamp coolers. And I've been in businesses north and south and almost all use AC in the summer.
ReplyDeleteThat's a ridiculous statement. All business use ac and most apartment buildings. The majority of single family homes also. You must not live where I have for 67 years.
DeleteCorrelation isn't necessarily causation.
ReplyDeleteWHO has attempted to correlate influenza with whatever Low humidity had some correlation. In temperate areas the wintertime heating dries out air and we are cooped up. In tropics maybe dry season or cooped up air con. No really high correlations. So I agree that recirculated air in crowded areas does spread respiratory viruses. The other data issue is listing someone as having flu or COVID. Often not diagnosed and maybe incorrectly. Tanzania used to report no flu deaths until they started testing for flu. Why? Because malaria with fever and sore muscles kills a lot of folks - so flu was and is undercounted. And final cause of death is often a guess. And often no autopsy. So Cliff thanks for looking at correlations. Maybe someday humans will understand more of the natural world. I will not quarrel over an interesting stab at how a totally new virus spreads. We all need to share ideas and look for relationships
ReplyDeleteI was wondering if anyone in the USA was going to figure out this rather obvious relationship of noticing how the daily temperature maps closely reflect the Covid case load. Congratulations to a very smart professor from my Alma Mater. Cheers from Whistler, BC where it is too cool for AC.
ReplyDeleteThis is suppose to be a weather blog not a left wing politics blog. Want that got to Reddit! They'll adore this kind of BS.
ReplyDeleteIf you want to see government control at it's most look up central heating Moscow and see why they have been dismantling it a piece at a time.
ReplyDeleteCliff is neither an epidemiologist or virologist, so perhaps he could stop playing like one. If he feels he has something meaningful to contribute, why doesn't he get on as a co-investigator on a project and get his name on a scientific paper. The UW and its affiliates have some of the best researchers in the world, but I am sure they could use additional expertise.
ReplyDeleteAlready looking forward to your post six months from now: Are furnaces and heat pumps contributing to covid-19 spread? It's not that people want to be inside out of the heat, it's that they're not wearing masks and social distancing when they do. The head of the CDC echoed what many have said in the past month: if even 80% of the world population wore masks all the time in public for 6 weeks, the virus would disappear from existence.
ReplyDeleteThe CDC has gotten pretty much everything wrong until now about this virus. Why should we trust them now?
DeleteIs there any way for you to disaggregate increases time spent indoors from AC use? With aerosolized particles, it really could be either or both. With a temp change coinciding with loosening of social guidelines, I think you need more detailed info before you can point to AC specifically. It is a powerful hypothesis though
ReplyDelete"The headlines are screaming about recent increases in coronavirus cases, with some suggesting that the essential problem is the loosening of the lockdowns and restrictions. A number of media sources note that many of the problematic locations are "red" states with Republican leadership."
ReplyDeleteThe media should try to not exclude important details - like the fact testing has dramatically increased therefore cases are up. The vast majority of whom are asymptomatic and will never be sick let alone need hospitalization.
You’re right as far as it goes — people need to avoid being inside except for their home.
ReplyDeleteI’m not here to condemn governors who rushed to reopen, but it’s probably worth noting that New Mexico is seeing fewer cases than Arizona and Texas. Likewise, Louisiana is in better shape than its neighbors. Let’s leave parties out of it: reopening early was a worse mistake in hotter states.
Is the simply about being in a confined space circulating the air being an ideal environment for the virus to spread?
ReplyDeleteI have been thinking the same thing about A/C but I suspect it is just one of many factors in play. Clearly non adherence to cdc protocol is a bigger one.
ReplyDeleteCliff,
ReplyDeleteIt makes common sense that air conditioning would cause more cases in states with extremely hot and humid weather. Also, it's customary for almost everyone to rely on air conditioning in those places. You go from your air conditioned vehicle to your air conditioned destination then back to your air conditioned home.
I hope you can use this information to work with others who can assess the level of problem this presents in different states, cities and communities. If it saved lives it would certainly be worthwhile. I'm so glad you invested your time and energy into looking at this portion of the Covid-19 crisis.
Most homes in dry places like Arizona are not using air conditioners. They use swamp coolers that draw outside air through a filter continuously soaked with water that uses evaporation to cool the air. This also increases its humidity. The increases in cases are due to people being together in businesses that reopened that are more likely to use traditional air conditioners, not being together at home.
ReplyDeleteNice post. I’ve been spamming threads with this theory for months. I now think it must be behavioral. Before the disease was very prominently up north:
ReplyDeletehttps://noconsensus.wordpress.com/2020/05/04/uncommon-cold-doesnt-break-the-rules/
I thought sunlight would help a lot, then cases boomed in the south but North looks pretty good. Now the north is starting to get worse. I now think it is because we are going indoors due to weather. Too cold or too hot, we do the same thing and transmission risks increase. Remember during the height of this, when NY reported that 66% of new cases were from quarantined people.
UV kills the virus, go outside.
I'd posit you'd get a similar correlation if you plotted % cases against % Republican lawmakers. Which in turn would look almost the same as % businesses open, and therefor % attendance (and mask-less social crowding) at businesses, beaches, etc.
ReplyDeleteCraig..you would be wrong. Look at a map of party affiliation (e.g.https://www.dailykos.com/stories/2017/1/30/1627319/-Daily-Kos-Elections-presents-the-2016-presidential-election-results-by-congressional-district). Many of the worse-off places are highly democratic (e.g. southern Arizona, large swaths of the SE)
DeleteCliff, according to the map you present, 8 of the 9 states have Republican governors. Admittedly this is driven by "correlation is not causation", but so is much of this article. Moreover, just because some of the worst counties are run by Democrats" does not mean that Republican government is not a driving force for some of the problem, as state governments are probably more powerful than local or the federal government in Covid 19 leadership. The counties you note that have high infection rates and Democratic local government, which are Predominately "Democratic", and have high infection rates, are also heavily populated by minorities, Latinos in particular. You can see this play out in Imperial Valley in California and Eastern Washington. Another words, this could be a problem of Multicollinearity (Which is probably going to be a problem with many variables) or a data set that is not detailed enough. Thus, it probably would have been beneficial to put a county map next to a state map, as this would provided more detailed information.
DeleteYou could also run a multivariate model and then evaluate the correlation coefficients and multicollinarity. Then you could put the political correlation coefficients on state and county maps, color them, and see how they line up to political leadership at the top of the state or at a local level. If you wanted to take perhaps a better snapshot of variables that probably have a more direct effect on virus spread, but is somewhat related to politics, look at the demographic characteristics age, race, mask wearing, and the nature of closings in a given state and county.
Which is one reason the spread is low here in the FLA Keys: We enjoy the great outdoors. I spend all day in the shade of my fig tree in my gym shorts in the breeze with a cold drink reading library books.
ReplyDeleteLove the Florida Keys...Key West above all...
DeleteThis claim has a major flaw. The focus is on increasing rates, so obviously states where the rates had been low are going to see the largest increases. It is the flip side of NYC, where the at-risk elderly were so poorly protected that rates are going down because there are so few at-risk people left.
ReplyDeleteAnd quite frankly, I have the same attitude toward this virus as the Israelis had toward terrorist bombings when I lived there. I'm not going to put my life on hold for endless months with repeated lockdowns because many of those in public health and politics have a Napoleonic complex, preferring highly intrusive and destructive solutions to medical drugs know to be effective from a daily Vitamin D pill to HCQ, which is probably as effective with preventing Covid-19 as it is with malaria. What they offer isn't the best solutions. What they offer are costly and ineffective ones.
Here's an analysis of a gold standard research trial published in the prestigious New England Journal of Medicine that shows that HCQ is at least 17% effective at preventing a high-risk exposure to this virus from becoming symptomatic Covid-19 even when given three or four days after exposure. Given quickly or taken as a preventative its probably more than 50% effective. Put in widespread use, we could achieve the equivalent of herd immunity in mere weeks. Click on the cloud icon in the lower right and you can download a PDF to send to others.
https://indd.adobe.com/view/e37cec42-6ca4-48d5-84b7-f7d0e373e305
Many of the A/C units or components are made in China. Just can't trust these folks at all.
ReplyDeleteAt a certain point, we take the risk and re-open the economy. You can't mask up everybody. Kids under 5 are exempted (rightfully) and they are still going to preschool. Just let the disease make its course.
ReplyDeleteBe thankful for a man like Cliff Mass, this information, real facts and studies is very hard to find and not available to most people. I appreciate his honest interpretations of data, even if it goes against the accepted theories.
ReplyDeleteMany researchers have been looking at temperature and relative humidity in connection with spread of the coronavirus. It looks like RH may be more important than temperature, with lower RH associated with higher transmission rates. Most indoor air control (whether AC or heating) reduces RH. But if you aren't sharing indoor space with people who could transmit the disease, you aren't going to get sick--so this remains about social distancing and mask use.
DeleteViruses only care about 1 thing, opportunity (well care is a bit strong when you are talking about sort of alive RNA microbodies). Oregon, where I am at, has had strong safe distancing requirements, regardless of bar or restaurant or wherever people gather. If you are in a place with AC, but wear masks and maintain distance, I doubt AC would play too much of a role compared to if you allow no masks and don't care about crowds. And the level correlation between outbreaks and crowds is much higher than any other metric.
ReplyDeleteI think the point that Cliff is making is not that AC ITSELF increases the spread of the virus, but rather that people in hot locations tend to gather in places with AC.
ReplyDeleteI just don't think that masks will do much in such places. People tend to move them up and down. And if you are eating or drinking you have to remove them. Then, there is the issue that a lot of people who move from a hot place to a cold (under-cooled) place end up ... sneezing.
And in the fall and winter and spring seasons here in the Pacific Northwest, people will once again be congregating indoors to avoid the cold and wet weather. So we can expect an increase in virus transmission here again too.
ReplyDeleteIn a building, most ac units are in there own room/ enclosure. They pull air from that room and disperse it to other rooms in the building. Most people walk by this room and don't even know it's there. Air is not pulled from one room full of people and pushed to another room of people. And the units have filters, probably not hepa,but they're dense. I would like to see some research on viruses making it through ac unit filters and ducts when the virus is in the ac unit enclosure, outside the enclosure, even down the hall. The better we understand how it can travel the better.
ReplyDeleteFor every complex problem there is an answer that is clear, simple, and wrong.
ReplyDeleteH. L. Mencken
SINGAPORE - From a WUWT article several days ago "Singapore’s difficulties with Covid could be due to factors which interfere with realisation of the potential benefits of Singapore’s tropical humidity. Singaporeans spend a lot of time indoors in air conditioned comfort, they take their air conditioning very seriously." https://wattsupwiththat.com/2020/07/12/covid-19-the-winter-disease-which-hates-humid-summers/
ReplyDeleteI think this ignores that people are going out partying, lockdown restrictions are more lax in these states than others, people are going to beaches in large groups...
ReplyDeleteGenerally speaking kids under 5 are not going to preschool (that struck me as a silly comment so I started there). Just had to say that. I like ya Cliff but your answer is speculative at best at this point. Without an ACTUAL study that controls for all other possible causes of transmission you really don't have a clue, well probably more of a clue than I but still... we all want definitive easy answers to everything. COVID-19 has a few. Airflow matters. Aerosolizing (spelling?) is now being found to be a HUGE factor. I can see air conditioning having a greater affect on aerosolizing. No doubt the simple facts (ah simple) that folks are congregating more, in closer proximity with no masks, provides ample opportunity for this little bugger to latch on and spread. Ya politics have seemingly influenced unwise decisions not to mask. There are so many factors that favor the virus to spread in those states that pointing to one simple thing does not mean one simple answer. The longer you stay in aerosolized air the more you are open to the virus invading. Airflow, airflow, airflow. Be kind to others and wear masks. Wash, wash, sanitize!!!
ReplyDeleteSorry, I didn't mean to write that all kids under 5 go to preschool. But a lot of kids, including under 5, in Seattle, yes, are still going to preschool or day camps. I'm not bashing parents who send their kids to preschool or camps. This thing with masking up is an illusion. And, for how long are we going to do it? Until there is a vaccine? Good luck.
DeleteI'd like to inform/remind folks here that there are good positivity models (yes, it's complicated enough to need a model) at https://rt.live/, with data freely available.
ReplyDeleteAlso, I recently stumbled across https://www.covidexitstrategy.org/, which is a really good compilation of data of all kinds... Though I'm pestering them about making their complete historical data more easily available.
Indoors versus outdoors. Pretty simple
ReplyDeleteThere might be just a little something to this, but it is obvious that the overwhelming factor in the increased cases is the disjointed leadership/message on a state by state basis, and complete lack of leadership on the national level. That coupled with many states re-opening and in some cases going back to business as usual, and this is what you get. Other countries around the world that are doing well have unified leadership and citizens who don't all believe everything is a hoax. And many of these countries have hot temperatures and homes/businesses with air conditioning. Let's spend less time on the minutia of hot vs. cool weather, air conditioning, etc, and more time focusing on the real problems this country is facing. No leadership, nobody on the same page equals chaos and a virus that will not go away.
ReplyDeleteAll public buildings and most commercial buildings in washington have have basically the same HVAC setup as any southern state. The amount of cooling varies based on need. The only real difference is where you spend your time; In an mall, bar or restaurant, or at home. People in any climate with any sense are not going to the mall, bars or restaurants. People led and believing in idiot leaders will gladly do so.
ReplyDeleteYou are right: looks like, feels like, 80 years ago = Nazis.
ReplyDeleteThanks! Neil goldberg