Last week the Washington Post published a fascinating article on the data visualisation work of the Donald Trump media campaign. In my last job I frequently harped on the importance of displaying the baseline and/or setting the baseline to zero. When you fail to do so you distort the data. But maybe that is the point of this, for lack of a better term, political data visualisation.
My favourite author is George Orwell of 1984 and Animal Farm fame. But Orwell also penned numerous essays, one of which has struck me as particularly relevant in this election cycle: Politics and the English Language. In concluding the essay Orwell wrote:
Political language…is designed to make lies sound truthful and murder respectable, and to give an appearance of solidity to pure wind.
And so political data visualisation? Well I believe it exists to serve the same purpose. The article goes into detail about how the designers behind the graphics fudged the numbers. Now did the campaign intend to mislead people with the data visualisation graphics? It is hard to say, because some of their graphics actually diminish leads that Trump has among certain demographics. Could it be the designer behind the graphics simply does not understand what he or she is doing? Perhaps. We clearly cannot know for certain.
Either way, it points to a need for more understanding of the importance and value of data visualisation in the political discourse. And then the natural follow-up of how to best design and create said visualisations to best inform the public.
But I highly recommend going to the Post and reading the entirety of the article.
Credit for the original work goes to the Trump campaign graphics department, the criticism to John Muyskens of the Washington Post.
In politics, it is really easy and often popular to bash the federal government. Especially when it comes to its penchant for collecting taxes to pay for things. And sometimes those things are in other states than your own. But do you know how much federal money goes back to your own state? Well now you can thank the Pew Charitable Trusts for putting together this piece that explores what percentage of state budgets is comprised of federal grant money.
While the piece also includes a donut chart—because why not?—my biggest gripe is with the choropleth and the choice of colour for the bins. If you look carefully at the legend, you will see how both the lowest and highest bins use a shade of blue. That means blue represents states that receive less than 25% of their budget from federal grants and also states that receive more than 40% of their budget from the same federal grants. But if your state is between 25% and 40%, your state suddenly turns a shade of green. It really makes no sense. I think the same colour, either blue or green, could be used for the entire spectrum. Or, if the designers really wanted a divergent scheme, they could have used the national average and used that as the breakpoint to show which states are above and which are below said average.
Credit for the piece goes to the Pew Charitable Trusts graphics department.
One of the things I like about Chicago’s WGN network is its weather blog. They often include infographic-like content to explain weather trends or stories. But as someone working in the same field of data visualisation and information design, I sometimes find myself truly confused. That happened with this piece last Friday.
The map in the upper-right in particular caught my attention and not in the good way. The overall piece discusses the heavy rainfall in the Chicago area on Thursday and the map looks at the percentage increase in extreme weather rainfall precipitation. All so far so good. But then I look at the map itself. I see blue and thing blue > water > rainfall. The darker/more the blue, the greater the increase. But, no—check out Hawaii. So blue means less rainfall. But also no, look at the Midwest and Southeast. So does green mean anything? Beyond being all positive growth, not that I can tell. As best I can tell, the colour means nothing in terms of rainfall data, but instead delineates the regions of the United States—noting of course they are not the standard US Census Bureau regions.
So here is my quick stab at trying to create a map that explains the percentage growth. I have included a version with and without state borders to help readers distinguish between states and regions.
And what is that at the bottom? A bar chart of course. After all, with only eight regions, is a map truly necessary especially when shown at such an aggregate level? You can make the argument that the extreme rainfall has, broadly speaking, benefitted the eastern half of the United States. But, personally speaking, I would prefer a map for a more granular set of data at the state or municipality level.
Credit for the piece goes to Jennifer Kohnke and Drew Narsutis.
Another day in Philadelphia, another post of Philly data visualisation work. Here we have a piece from 2015 that was updated earlier this spring. It looks at overdose rates in the Philadelphia region, including parts of New Jersey. It does include a map, because most pieces like this typically do. However, what I really find interesting about the piece is its use of small multiple line charts below to take a look at particular counties.
The piece overall is not bad, and the map is actually fairly useful in showing the differences between Jersey and Philadelphia (although why New Jersey is outlined in black and the Philly suburban counties are not I do not know). But I want to take a look at the small multiples of the piece, screenshot below.
You can see an interesting decision in the choice of stacked line charts. Typically one would compare death rates like for like and see whether a county is above, below, or comparable to the state, local, or national averages. But combining the three gives a misleading look at the specific counties and forces the user to mentally disentangle the graphic. I probably would have separated them into three separate lines. And even then, because of the focus on the counties, I would have shifted the colour focus to the specific counties and away from the black lines for the national average. The black is drawing more attention to the US line than to the county line.
I am on holiday for a few days and am visiting Philadelphia. So what better time to cover some Philadelphia-made content? This interactive piece came out last year from Philly.com alongside coverage of the Philadelphia mayoral contest.
I want to call out the colour palette for the choropleth in particular. We can see a blue to red system with a stop at yellow in the middle—a divergent palette. With this kind of a setup, I would expect that yellow or the light blue to be zero or otherwise straddle the point of divergence. Instead we have dark blue meaning 0 and dark red meaning 401+. The palette confuses me. It could be that the point of divergence—something around the 200 number—could be significant. It could be the city average, an agreed upon number for good neighbourhood relations, or something. But there is no indication of that in the graphic.
Secondly the colour choice itself. I often hesitate using red (and green) because of the often-made Western connotation with bad. Blue here, it works very well with the concept of the thin blue line, NYPD blue, blue-shirted police. If we assume that there is a rationale for the divergent palette, I would probably place the blue on the high-end of the spectrum and a different colour at the negative end.
Lastly, from the perspective of the layout, Philly has a weird shape. And so that means between the bar chart to the right and the city map on the left the piece contains an awkward negative space. The map could be adjusted to make better use of the space by pointing north somewhere other than up.—why is north up?—to align the Delaware River with the bars. Or, the bars could abut West Philly.
The interactions, however, are very smooth. And a nice subtle touch that orients the reader without distracting them is the inclusion of the main roads, e.g. Broad Street. The white lines are sufficiently thin to not distract from the overall piece.
And by this title I am not referencing McKinleys, K2s, or Everests. No, the BBC published this piece on the changing average heights of citizens of various countries. This was the graphic they used from the report’s author.
Personally speaking, I do not care for the graphic. It is unclear and puts undue emphasis on the 1914 figure by placing the illustration in the foreground as well as in the darkest colour. I took a thirty-minute stab at re-designing the graphic and have this to offer.
While I admit that it is far from the sexiest graphic, I think it does a better job of showing the growth than decline of national heights by each sex in each of these three select countries. Plus, we have the advantage of not needing to account for the flag emblems. Note how the black bars of Egypt disappear into the black illustration of the person.
Credit for the piece goes to the eLife graphics department.
I mean really, given the rampant and pervasive nature of the Russian state-aided doping programme, how could I not use the Russian reversal? Yesterday WADA, the international anti-doping agency, released its findings on Russian doping at the Olympics. And, suffice it to say, the report is rather damning. The BBC published this graphic in an article to help demonstrate the scheme.
Unlike the evidence of doping, I find the graphic itself lacking. More could have been done to create more consistent type. Text justification ranges (pun intended) from left to right, without any clear system. Why do some stages, e.g. four, align to the right and then others, e.g. seven, align to the left?
Also, I believe more could have been done with the illustrations, in particular the bottles labelled A and B, to better differentiate between a clean sample and a contaminated sample. Why, for instance, does Step 1 include both an A and a B when it mentions only one sample?
In short, the story certainly warrants explanatory graphics, especially as to how the sealed lids were removed, but this piece is not the solution (pun also intended).
Credit for the piece goes to the BBC graphics department.
Last one of these critiques—I promise. Earlier this week I looked at the New York Times’ coverage and the BBC’s coverage. Well, today I want to examine the Guardian’s coverage of the Brexit vote results. This piece differs the most from the preceding work and it starts right from the top, literally.
I am not the biggest fan of the illustrations of David Cameron and Boris Johnson, but in a sense, neither is a throw-in. For the last few months, the Guardian has been using these and similar illustrations of US presidential candidates to tie results into different political camps. Thus in that sense, they do fit the Guardian’s current brand. Interestingly, neither remains (pun intended) in the picture for the future of the Tories.
Data-wise, however, the decision to use the bar chart at the outset of the piece reflects an understanding of the importance of the top-line number. Districts count, but only at that granular level I discussed. What truly matter, though, is the aggregate. And this is a no-doubt-about-it means of conveying that information. (I will admit the David Cameron frowny face does help a wee bit.)
And if the use of big numbers and illustrations at the top of the piece broke with the choropleth map we saw with the New York Times and the BBC, well, we have another clear break.
Instead of using a geographic map, the Guardian employs a cartogram with hexagons. I have covered similar uses a severaltimesbeforetoday. The hexagon shape allows better retention of familiar geographic shapes, while still providing a means of solving the small district problem, especially in places like central London.
From another design perspective, that of colour, we see an improvement over the blue–yellow spectrum used by the BBC. You may recall from yesterday:
Having multiple tints and shades of yellow makes the map difficult to read.
Here, the Guardian instead opted for a simplified, and easier to read, two-step split. Bright blue and yellow with each have a call it half-tint. With only two blues and, more importantly, two yellows to distinguish, the map becomes easier to read. The trade-off, the darker of the colours represents anything above a 15% majority.
Clicking on the map then provides with a small summary of the district results.
Here we see nothing too dissimilar from how the BBC treated the interaction with their map. A small, subtle design element I enjoy, however, is the inclusion of the national average. The 50% marker indicates clearly which side won, but the tick below the bar gives the reader context of where the district fell into relation to the remainder of the country.
And that leads us into the next set of comparisons.
The Guardian took local district results and compared them against several different demographic and socio-economic indicators. This allowed them to present various correlations of the vote. It turns out that higher education correlated best with the results of the UK vote. From a design perspective, the linked circles provides some stability. However, I would have preferred the ability to click a geography and have it remain sticky and bring up the specific figures. Additionally, some sort of text search for geographies would be helpful.
And then the Guardian’s piece closes as strongly as it opened.
The piece examines three riverside areas to provide specific analysis to the vote. The screenshot above focuses on the Tyne, which runs alongside the aptly named Newcastle upon Tyne. The Guardian uses the previous general election results for the area to contrast with the referendum results. It does similar analysis for the Thames (London) and the Mersey (Liverpool).
Similar to the New York Times piece, the Guardian’s piece responds well to viewing the content on a small screen. The changes are less complex and they deal mostly with the arrangement of the various components instead of the layout of contextual data. But the Guardian clearly considered how the piece would work on a mobile phone up through a widescreen monitor.
Overall the piece is quite strong and does an excellent job of showcasing the results data and providing insightful analysis that complements the vote totals.
Credit for the piece goes to the Guardian graphics department.
So now it is two weeks since the Brexit vote. Yesterday, I looked at the results designs from the New York Times. Today I want to take a look at those of the BBC. Not surprisingly the two share in the use of choropleth maps; the choice makes a lot of sense. People vote within districts and those form the most granular unit of data available. But, whereas the New York Times led and really focused on one giant map, the BBC opted to use multiple, smaller maps. (They did choose a different page for their live results, but we are comparing post-result coverage.) For example, their piece leads in with a map of Leave’s results share.
There are a few key differences between this and the New York Times. First and foremost, this map is interactive. Mousing over various districts provides you the name, and by clicking you move into a zoomed-in view of the district. It displays the district name, the vote totals and share for the two camps, and then voter turnout. From a design standpoint, the problem with the zooming in is that the scales of the outlining stroke does not change.
A thin stroke at the national, zoomed-out view, translates to a thick, clunky, and awkward-looking outline at the local, zoomed-in view. And as the above screenshot highlights, many of the urban districts are small in comparison to the more rural districts. Unfortunately the map does not offer the functionality of zooming-in prior to selecting a district. So many of the districts in the more urban areas like London, Manchester, Birmingham, and Belfast are difficult to see and select. Thankfully, below the map the BBC offers a function to type in your district, post code, or Northern Irish constituency to help you find smaller districts.
Another design criticism I have with the piece is the colour palette. Broadly speaking, the piece uses blue and yellow. The two colours make sense in a few ways. Both are present on the European Union flag, with yellow stars on a blue field. (Importantly the twelve stars do not represent EU members like the US flag’s fifty stars represent the states.) Another, far looser interpretation could be the blue of the Conservatives and the yellow closer to the gold of the UK’s Liberal Democrats, the former broadly anti-EU and the latter pro-EU. Regardless of the rationale, the choice of yellow to display multiple levels of data is less than stellar (pun intended), as this Remain share map highlights.
Having multiple tints and shades of yellow makes the map difficult to read. The lowest value yellow is brighter than the next higher level, and so stands out more vividly on the map than those districts that had a higher share of Remain votes. Using yellow against blue does work, especially in the bar charts throughout the piece and seen in the aforementioned Islington screenshot. But, as a colour for wider, more intense use, yellow was not the wisest decision.
The BBC also included several other choropleth maps exploring the vote breakdown. In this instance of voter turnout, we have the same choropleth map, but a green colour indicating the total vote turnout.
The colour and its choice makes broad sense; green is what one gets when they mix yellow and blue, when you combine Remain and Leave. However, the map functionality of clicking to reveal results still shows the overall results.
At this point, we have moved on from the vote results themselves to the breakdown of the vote. I would have redesigned the mouse-click to display a results view that highlighted turnout over the results themselves. Certainly keeping the results is important, but the focus of this map is not the vote, but the turnout. The data display should be designed to keep that consistent.
One part of the piece where I quibble with the designer selection of chart type follows on from turnout: a comparison of turnout to the youth population.
Asking people to compare undistinguished districts on one map to those of another—note the white district lines have here disappeared—is difficult. My first thought: I would have instead opted for an interactive scatterplot. Comparing the turnout on one axis and youth on the other, the user would have an easier time identifying any correlations or clusters of data.
In contrast, the following map comparison would not work via a scatterplot. Here we compare June’s results to those of a vote in 1975. In the intervening years, the geography of the voting districts changed, and so a one-to-one comparison is impossible.
The broad scope, however, is clear. A resounding vote to stay part of the European Market or single market in 1975 evolved into a narrow but decisive vote to leave the European Union in 2016.
The piece then closes out with an interactive map of the total results and then, importantly, a long list of bar charts showing each district’s results. Unlike the map, however, the bar charts are a static graphic. And with a few hundred to view, it becomes difficult to isolate and compare two in particular. But the selection of the visualisation type makes a user’s comparison far more precise.
Overall, I would rate the piece a solid work, but with some clear areas of improvement. And who knows? Maybe there will be a second referendum. Or a new general election. And in that case, the BBC could improve upon the designs herein.
Credit for the piece goes to the BBC graphics department.
Well a little under two weeks later and here we are: Brexit. I wanted to take a moment in a slightly longer piece and comment on it. Not the results, because no, that I can leave to a pint at the pub. Instead I wanted to comment on this particular results content from the New York Times that I rather admire.
Overall the piece is not interactive; it features a static choropleth map with annotations and insets, particularly of greater London. On a side note, I would be remiss if I did not point out that similarly to the piece I wrote last week, this map omits a voting district: Gibraltar. Gibraltar, like Northern Ireland, borders the European Union directly via Spain. And despite voting overwhelmingly to remain in the EU, Gibraltar is omitted from these results.
In a large layout, the piece makes excellent use of annotation text to indicate the overview stories for the home nations of the United Kingdom: Scotland, Wales, and Northern Ireland. Northern Ireland, of course, will likely have to deal with the reintegration of border controls between the Republic of Ireland and Northern Ireland, a point the piece makes quite clear.
Additionally the map makes use of small elements to draw attention to data points, i.e. geographies, worth noting. London dominates the urban landscape, but other important cities like Belfast, Manchester, Birmingham are circled to show the strength of Leave/Remain. I would be curious to know the rationale behind including some areas, but omitting others, e.g. the strong Remain results in Cambridge or Brighton or the strong Leave results in Boston, require knowing just where cities are located in England.
From a design standpoint, the colours used in the map work really well together in contrast to other palette choices one could make. (We will take a look at that tomorrow.) Additionally, the shape of the United Kingdom allows for contextual elements, e.g. the regional result aggregates, to be placed much closer and nearer to the results. The space also allows for those annotations to be placed near their particular geographies.
But, what makes the piece stand out is when the user consumes it on smaller screens. On a more tablet-sized screen, we see a tweaked layout.
It makes use of the remaining wide-aspect dimensions to move the greater London results into a white space carved out by the peninsula of East Anglia. While the city and home nation labels remain, the regional annotations and results are gone from the graphic. Instead, they have been placed below the map, the main and most important part of the story.
Then for mobile phone or other narrow displays, the piece degrades even further.
City labels and circles are gone, with the exception of London. The greater London inset moves from alongside the map to now below the map, in the Channel so to speak. This layout allows for a narrow screen to better view the geographic results and then scroll down into the districts of London that require more space to be displayed. The annotations and stories remain below the graphic.
The design of the overall piece accounts nicely for at least three different screen sizes while keeping the story constant. All the truly changes is the layout of the graphic (and the loss of a few contextual labels at the smallest of sizes). Overall, it makes for a rich and compelling—and well designed—piece on the Brexit results.
Credit for the piece goes to Gregor Aisch, Adam Pearce, and Karl Russell.