As I said yesterday, since people are finding these updates helpful on the social media, I am going to repost the previous evening’s graphics I make on the Coronavirus Covid-19 outbreak here on Coffeespoons as well. So while today is Thursday, these are the numbers states provided yesterday, so it’s more of a Wednesday update.
But here I can start with the flatter curves graphic. The New Jersey numbers in particular look good—I mean they’re still bad. Of course we are just a few big breaches of quarantine and lapses in social distancing from reversing that progress.
Maybe some curve flattening?
State-wise, Pennsylvania continues to worsen. However, a close look at the slope of the line in the previous chart indicates that the steepness of the growth may be lessening. Deaths passed 300 and cases are now firmly entrenched on both sides of the state with the rural, less densely populated areas in the Ridge and Valley portion of the state seemingly hit not as hard.
The situation in Pennsylvania
Despite the potential flattening, New Jersey is just in a rough spot. The final bastions of low case numbers in South Jersey are slowly filling up as Cape May County passed the 100-case threshold.
The situation in New Jersey
Delaware continues to accelerate and is now past 1000 cases.
The situation in Delaware
Virginia continues to see cases spreading in the eastern, more populous portions of the state. And at 75 deaths, it’s nearing the 100-death threshold.
The situation in Virginia
Illinois is seeing deaths occur away from Chicago, in the St. Louis suburban counties and in and around Springfield and Champaign and Bloomington areas.
Here are the Tuesday figures for Pennsylvania, New Jersey, Delaware, Virginia, and Illinois. At the end is an updated version of the flattening curves chart as well. Given the value of these graphics that people have been texting, emailing, and DMing me on social media, I might consider making these a regular staple here on my blog as well. I would probably slowly write about other graphics covering the outbreak as well.
Any feedback is welcome on how to make the graphics more useful to you, the public.
Pennsylvania has finally reached the point where the virus has infected at least one person in every county. Now, if we shift our attention a wee bit to the deaths, we can see those are still largely confined to the eastern third of the state.
The condition in Pennsylvania
New Jersey continues to suffer greatly. But a sharp increase in new cases could be a blip, or it could mean the curve isn’t flattening. We need more data to see a longer trend. Regardless, over 3000 more people were reported infected and over 200 more died.
The condition in New Jersey
Delaware worsened significantly. As a small state, it has a lower captive population. But it is rapidly approaching 1000 cases. In fact, I would not be surprised if that is the headline from Wednesday.
The condition in Delaware
Virginia also saw a significant uptick in cases. And most counties and independent cities in eastern Virginia now report cases. But the rural, mountainous counties in the west and southwest are not uniformly infected. At least not yet.
The condition in Virginia
Illinois saw some geographic spread, but again, compared to a state like Pennsylvania, the worst in Illinois is disproportionately concentrated in the Chicago metropolitan area.
The condition in Illinois
Lastly, the curves are not flattening in all the states but maybe New Jersey. But as I noted above, the higher daily cases there might be a blip.
Yesterday I mentioned that there was some data to suggest that at least in New Jersey the curve was flattening. Monday we received additional data and so I wanted to share what that data showed.
I used a set of bar charts to show the new daily cases yesterday for Pennsylvania, New Jersey, Delaware, Virginia, and Illinois. But as I mentioned, a single day can be a blip. Noise. We want to find the pattern or the signal within that data set. Consequently I applied a simple 7-day rolling average to the new daily cases data set.
I chose seven days for two reasons. The first was that the onset of the symptoms is 5–10 days after infection. Picking a mid-point in that range assures us that people who are infected are beginning to appear in the data. Secondly, a cursory check of the data suggests that reported numbers dip lower on weekends. And so making a week-long average covers any possibility of lower values at week’s end.
That preface out of the way, what do we see? Well, there is some evidence that the curve is flattening in New Jersey. The lines below represent that rolling average. And if you look at the very top of the New Jersey curve, you can see it beginning to flatten.
Flattening curves
Unfortunately that does not mean New Jersey is out of the woods. Not by a long shot. Instead, that means tens of thousands of people will still be infected. And hundreds more will die. But, the rate at which those two things happen will be lower. Hopefully hospitals will not be as overwhelmed as they presently are. And that might make for a lower total death count.
The data does not support, however, the notion that the curve is flattening in the other states. Consider that the United States spans a continent and contains over 330 million people. The outbreak will look different in different states. Compare Pennsylvania and Illinois, which have similar case numbers. But in Pennsylvania we have more cases in smaller cities and rural areas and fewer in the largest cities. Plus, of course, we have the different measures taken by different states to contain and mitigate the pandemic within their borders.
But, we do have some data to suggest that at least in New Jersey the curve is flattening. I’ll take good news where I can find it. (Even if it comes from Jersey.)
This past weekend I continued looking at the spread of COVID-19 across the United States. But in addition to my usual maps of Pennsylvania, New Jersey, Delaware, Virginia, and Illinois, I also looked at the number of cases across the United States adjusted for population. I then looked at the five aforementioned states in terms of new cases to see if the curve is flattening. Finally, I looked at the number of hospital beds per 1000 people vs the number of cases per 1000 people.
The latter in particular I wanted to be an examination of hospitalisation rates vs ICU beds, which are a small fraction of total hospital beds. But as I could not find that data, I made do with overall cases and overall beds.
So first let’s look at the cases across the U.S. What you can see is that whilst New York and New Jersey do have some of the worst of the impact, Washington is still not great and Louisiana and Michigan are also suffering.
The situation across the United States
And then when we look at the states by their cases per 1000 people and their hospital beds per 1000 people, we see that the states often claimed to be overwhelmed, New York, New Jersey, and Washington are all well over the blue line, which indicates an equal number of beds and cases per 1000 people, or near it. Because it is important to remember that not all beds are the type needed for COVID-19 victims, who often require the more fully kitted out ICU beds. Additionally, not all cases are severe enough to warrant hospitalisation.
Cases per 1k people vs hospital beds per 1k people
Then from the broader national view, we can look at the states of interest. Here, those of you who have been following my social media posts, you can see fewer dark purples in these maps. That’s because I have adopted a new palette that has sacrificed granularity at the lower end of the scale and added it at the top, a particular need in New Jersey and the Philadelphia and Chicago metro areas. And finally we look at the daily new cases to see if that curve is flattening.
Pennsylvania now has almost every county infected. But unlike Illinois, which has a similar infection rate but more unaffected counties, Pennsylvania has fewer cases in its big city, Philadelphia, and has more cases in the smaller cities and towns.
The situation in Pennsylvania
New Jersey is just a disaster. Deaths are now reported in every county—so I can probably remove those orange outlines. The only potential good news is that new cases for the second day in a row were fewer than the day before. It could be a blip. But it could also be a signal that the peak of infection has or is nearing. That said, hospitalisations and deaths are lagging indicators and could take two weeks to follow the positive test results. So in the best case scenario that this is a peak, New Jersey is far from out of the woods.
The situation in New Jersey
Delaware is the smallest state I look at—and one of the smallest in the union overall—but its cases are worryingly increasing rapidly, although like every state I examine in detail it had fewer new cases Sunday than Saturday.
The situation in Delaware
Virginia is in a better spot overall than the other four states. You can see that in the national map above. And most of Virginia’s cases are concentrated in the DC and Richmond areas as well as the cities along the peninsulas jutting into the Chesapeake.
The situation in Virginia
Illinois is, as noted above, similar to Pennsylvania in terms of infections. In terms of deaths, however, it is doubling Pennsylvania’s numbers. And most of its cases are located in and around Chicago. Big chunks of downstate Illinois are unaffected or lightly affected compared to the Commonwealth.
The situation in Illinois
Finally, as I noted in New Jersey, could these lower numbers Sunday than Saturday be meaningful? Possibly. But in all five states? Highly unlikely. Regardless, we can look at the number of daily new cases and see if that curve of infection is flattening. We should wait several days before beginning to make that assessment. But one can hope.
The case for flattening curves
All of this is to say that things are bad and likely will continue to get worse. But I will keep looking at the data daily and presenting it to the public to keep them informed.
By now we have probably all seen the maps of state coverage of the COVID-19 outbreak. But state level maps only tell part of the story. Not all outbreaks are widespread within states. And so after some requests from family, friends, and colleagues, I’ve been attempting to compile county-level data from the state health departments where those family, friends, and colleagues live. Not surprisingly, most of these states are the Philadelphia and Chicago metro areas, but also Virginia.
These are all images I have posted to Instagram. But the content tells a familiar story. The outbreaks in this early stage are all concentrated in and around the larger, interconnected cities. In Pennsylvania, that means clusters around the large cities of Philadelphia, Pittsburgh, and Harrisburg. In New Jersey they stretch along the Northeast Corridor between New York and Trenton (and along into Philadelphia) and then down into Delaware’s New Castle County, home to the city of Wilmington. And then in Virginia, we see small clusters in Northern Virginia in the DC metro area and also around Richmond and the Williamsburg area. Finally in Illinois we have a big cluster in and around Chicago, but also Springfield and the St. Louis area, whose eastern suburbs include Illinois communities like East St. Louis.
19 March county wide spread of COVID-1919 March county wide spread of COVID-1919 March county wide spread of COVID-1919 March county wide spread of COVID-1919 March county wide spread of COVID-19
I have also been taking a more detailed look at the spread in Pennsylvania, because I live there. And I want to see the rapidity with which the outbreak is growing in each county. And for that I moved from a choropleth to a small multiple matrix of line charts, all with the same fixed scale. And, well, it doesn’t look good for southeastern Pennsylvania.
County levels compared
Then last night I also compared the total number of cases in Pennsylvania, New Jersey, Delaware, and Virginia. Most interestingly, Pennsylvania and New Jersey’s outbreaks began just a day apart (at least so far as we know given the limited amount of testing in early March). And those two states have taken dramatically different directions. New Jersey has seen a steep curve doubling less than every two days whereas Pennsylvania has been a bit more gradual, doubling a little less than every three.
State levels since early March
For those of you who want to continue following along, I will be looking at potential options this coming weekend whilst still recording the data for future graphics.
The British election campaign is wrapping up as it heads towards the general election on Thursday. I haven’t covered it much here, but this piece from the BBC has been at the back of my mind. And not so much for the content, but strictly the design.
In terms of content, the article stems from a question asked in a debate about income levels and where they fall relative to the rest of the population. A man rejected a Labour party proposal for an increase in taxes on those earning more than £80,000 per annum, saying that as someone who earned more than that amount he was “not even in the top 5%, not even the top 50”.
The BBC looked at the data and found that actually the man was certainly within the top 50% and likely in the top 5%, as they earn more than £75,300 per annum. Here in the States, many Americans cannot place their incomes within the actual spreads of income. The income gap here is severe and growing. But, I want to look at the charts the BBC made to illustrate its points.
The most important is this line chart, which shows the income level and how it fits among the percentages of the population.
Are things lining up? It’s tough to say.
I am often in favour of minimal axis lines and labelling. Too many labels and explicit data points begin to subtract from the visual representation or comparison of the data. If you need to be able to reference a specific data point for a specific point on the curve, you need a table, not a chart.
However, there is utility in having some guideposts as to what income levels fit into what ranges. And so I am left to wonder, why not add some axis lines. Here I took the original graphic file and drew some grey lines.
Better…
Of course, I prefer the dotted or dashed line approach. The difference in line style provides some additional contrast to the plotted series. And in this case, where the series is a thin but coloured line, the interruptions in the solidity of the axis lines makes it easier to distinguish them from the data.
Better still.
But the article also has another chart, a bar chart, that looks at average weekly incomes across different regions of the United Kingdom. (Not surprisingly, London has the highest average.) Like the line chart, this bar chart does not use any axis labels. But what makes this one even more difficult is that the solid black line that we can use in the line charts above to plot out the maximum for 180,000 is not there. Instead we simply have a string of numbers at the bottom for which we need to guess where they fall.
Here we don’t even a solid line to take us out to 700.
If we assume that the 700 value is at the centre of the text, we can draw some dotted grey lines atop the existing graphic. And now quite clearly we can get a better sense of which regions fall in which ranges of income.
We could have also tried the solid line approach.
But we still have this mess of black digits at the bottom of the graphic. And after 50, the numbers begin to run into each other. It is implied that we are looking at increments of 50, but a little more spacing would have helped. Or, we could simply keep the values at the hundreds and, if necessary, not label the lines at the 50s. Like so.
Much easier to read
The last bit I would redo in the bar chart is the order of the regions. Unless there is some particular reason for ordering these regions as they are—you could partly argue they are from north to south, but then Scotland would be at the top of the list—they appear an arbitrary lot. I would have sorted them maybe from greatest to least or vice versa. But that bit was outside my ability to do this morning.
So in short, while you don’t want to overcrowd a chart with axis lines and labelling, you still need a few to make it easier for the user to make those visual comparisons.
Credit for the original pieces goes to the BBC graphics department.
A few weeks ago we said farewell to John Bercow as Speaker of the House (UK). Whilst I covered the election for the new speaker, I missed the opportunity to post this piece from the BBC. It looked at Bercow’s time in office from a data perspective.
The piece did not look at him per se, but that era for the House of Commons. The graphic below was a look at what constituted debates in the chamber using words in speeches as a proxy. Shockingly, Brexit has consumed the House over the last few years.
At least climate change has also ticked upwards?
I love the graphic, as it uses small multiples and fixes the axes for each row and column. It is clean, clear, and concise—just what a graphic should be.
And the rest of the piece makes smart use of graphical forms. Mostly. Smart line charts with background shading, some bar charts, and the only questionable one is where it uses emoji handclaps to represent instances of people clapping the chamber—not traditionally a thing that happens.
Content wise it also nailed a few important things, chiefly Bercow’s penchant for big words. The piece did not, however, cover his amazing sense of sartorial style vis-a-vis neckties.
Overall a solid piece with which to begin the weekend.
Credit for the piece goes to Ed Lowther & Will Dahlgreen.
Last we looked at the revenge of the flyover states, the idea that smaller cities in swing states are trending Republican and defeating the growing Democratic majority in big cities. This week I want to take a look at something a few weeks back, a piece from CityLab about the elections in Virginia, Kentucky, and Mississippi.
There’s nothing radical in this piece. Instead, it’s some solid uses of line charts and bar charts (though I still don’t generally love them stacked). The big flashy graphic was this, a map of Virginia’s state legislative districts, but mapped not by party but by population density.
Democrats now control a majority of these seats.
It classified districts by how how urban, suburban, or rural (or parts thereof) each district was. Of course the premise of the article is that the suburbs are becoming increasingly Democratic and rural areas increasingly Republican.
But it all goes to show that 2020 is going to be a very polarised year.
Just before Halloween, NBC News published an article by political analyst David Wasserman that examined what airports could portend about the 2020 American presidential election. For those interested in politics and the forthcoming election, the article is well worth the read.
The tldr; Democrats have been great at winning over cosmopolitan types in global metropolitan areas in the big blue states, e.g. New York and California. But the election will be won in the states where the metropolitan areas that sport regional airports dominate, i.e. Pennsylvania, Michigan, Wisconsin, and North Carolina. And in those districts, support for Democrats is waning.
The closing line of the piece sums it up nicely:
…to beat Trump, Democrats will need to ask themselves which candidates’ proposals will fly in Erie, Saginaw and Green Bay.
But what about the graphics?
We have a line chart that shows how support for Democrats has been increasing amongst those in the global and international airport metros.
Democrats aren’t performing well with the non-global and international types of metros
It uses four colours and I don’t necessarily love that. However, it smartly ties into an earlier graphic that did require each series to be visualised in a different colour. And so here the consistency wins out and carries on through the piece. (Though as a minor quibble I would have outlined the MSA being labelled instead of placing a dot atop the MSA.)
A lot of these global metros are in already blue states
The kicker, however is one of those maps with trend arrows. It shows the increasing Republican support by an arrow anchored over the metropolitan area.
Lot of Trump support in the battleground states
The problem here is many-fold. First, the map is actually quite small in the overall piece. Whereas the earlier maps sit centred, but outside the main text block, this fits neatly within the narrow column of text (on a laptop display at least). That means that these labels are all crowded and actually make it more difficult to realise which arrow is which city. For example, which line is Canton, Ohio? Additionally with the labels, because they are set in black text and a relatively bolder face, they standout more than the red lines they seek to label. Consequently, the users’ focus falls not on the lines, but actually on the labels—the reverse of what a good graphic should do.
Second, length vs. angle. If all lines moved away from their anchor at the same angle, we could simply measure length and compare the trending support that way. However, it is clear from Duluth and Green Bay that the angles are different in addition to their sizes. So how does one interpret both variables together?
Third, I wonder if the map would not have been made more useful with some outlines or shading. I may know what the forthcoming battleground states are. And I might know where they are on a map. But Americans are notorious for being, well, not great when it comes to geography. A simple black outline of the states could have been useful, though it in this design would have conflicted with the heavy black labelling of the arrows. Or maybe a purple shading could have been used to show those states.
Overall, the piece is well worth a read and the graphics generally help tell the narrative visually. But that final graphic could have used a revision or two.
Credit for the piece goes to Jiachuan Wu and Jeremia Kimelman.
John Bercow is no longer the British Speaker of the House. He left office Thursday. Fun fact: it is illegal for an MP to resign. Instead they are appointed to a royal office, in Bercow’s case the Royal Steward of the Manor of Northstead, that precludes them from being an elected MP. Consequently the House of Commons then had to elect a new Speaker.
For my American audience, despite the same title as Nancy Pelosi, John Bercow had a very different function and came to it in a very different fashion. First, the position is politically neutral. Whoever the House elects resigns from his or her party (along with his or her three deputies) and the political parties abide by a gentlemen’s agreement not to contest the seat in general elections. (The Tories were so displeased with Bercow they were actually contemplating running somebody in the now 12 December election to get rid of him.) Consequently, the Speaker (and his or her deputies) do note vote unless there is a tie. (Bercow actually cast the first deciding vote by a speaker since 1980 back in April.)
Because the position is politically neutral, all MPs vote in the election and debate is chaired by the Father of the House, the longest continuously serving MP in the House. Today that was Ken Clarke, one of the 21 MPs Boris Johnson booted from the Tory party for voting down his No Deal Brexit and who is not standing in the upcoming election. The candidates for Speaker must receive the vote of 50% of the House. And so they are eliminated in successive votes until someone reaches 50% of the total votes cast, though not all MPs cast votes, since some have already started campaigning. (Today there were 562, 575, 565, 540 votes per round.)
Notably, today’s vote occurs just days before Parliament dissolves prior to the 12 December election. Bercow, who chose to retire on 31 October, essentially ensured that the next Parliament will have a Speaker not chosen what could well likely be a pro-No Deal Brexit, one of the things which the Tories have against him.
So all that said, who won? Well I made a graphic for that.
A very different accent will occupy the big green chair.
