I am a graphic designer who focuses on information design. My day job? I am the data visualisation manager for the Federal Reserve Bank of Philadelphia. (This blog is my something I do on my own time and does not represent the views of the Fed, blah blah blah legal stuff.) And with my main interest in information design—be it in the shape of clear charts, maps, diagrams, or wayfinding systems—I am fortunate that my day job focuses on data visualisation. Outside of work, I try to stay busy with personal design work. Away from the world of design, I enjoy cooking and reading and am interested in various subjects from history and geography to politics to science to the arts. And I allow all of them to influence my work.
A few weeks back now the Economist posted a graphic about the link between lead, silver, and the rise and fall of the Roman Empire. But not in the way you probably think. Instead, they graph the appearance of lead deposits in the glaciers of Greenland.
For the full explanation you should read the short article. But this piece was right up my alley. We have ancient history, economics, science, and a timeline. And all in one neat little chart.
Credit for the piece goes to the Economist’s Data Team.
We are inching ever closer to the US midterm elections in November. In less than a week the largest state, California, will go to the polls to elect their candidates for their districts. So late last week whilst your author was on holiday, the Economist released its forecast model for the results. They will update it everyday so who knows what wild swings we might see between now and the election.
I will strike out against the common knowledge that this is a wave election year and Democrats will sweep swaths through Republican districts in an enormous electoral victory. Because while Democrats will likely win more overall votes across the country, the country’s congressional districts are structurally designed to favour Republicans as a result of gerrymandering after the 2010 Census redistricting. The Economist’s modelling handles this fairly well, I think, as it prescribes only a modest majority and gives that likelihood as only at 2-in-3. (This is as of 30 May.)
But how is it designed?
The big splashy piece is an interactive map of districts.
It does a good job of connecting individual districts to the dots below the map showing the distribution of said seats into safe, solid, likely, leaning, and tossup states. However, the interactivity is limited in an odd way. The dropdown in the upper-right allows the user to select any district they want and then the district is highlighted on the map as well as the distribution plot below. Similarly, the user can select one of the dots below the map to isolate a particular district and it will display upon the map. But the map itself does not function as a navigation element.
I am unsure why that selection function does not extend to the map because clearly the dropdown and the distribution plot are both affecting the objects on the map. Redeeming the map, however, are the district lines. Instead of simply plopping dots onto a US state-level map, the states are instead subdivided into their respective congressional districts.
But if we are going so far as to display individual districts, I wonder if a cartogram would have been a better fit. Of course it is perfectly plausible that one was indeed tried, but it did not work. The cartogram would also have the disadvantage of, in this case, not exhibiting geographically fidelity and thus being unrecognisable and therefore being unhelpful to users.
Now the piece also makes good use of factettes and right-left divisions of information panels to show the quick hit numbers, i.e. how many seats each party is forecast to win in total. But the map, for our purposes, is the big centrepiece.
Overall, this is solid and you better bet that I will be referencing it again and again as we move closer to the midterms.
Credit for the piece goes to the Economist Data Team.
On Saturday Ireland announced the results of a referendum on changing its constitution to remove Article 8, which had made abortion illegal except in the case of risk of death to the mother. And that was it, none of the usual rape or incest clauses. I want to look at a little coverage of the results and we will start with the Irish Times.
Their presentation is straightforward, a parliamentary-like slider and a small choropleth. All the colours link to each other and you will note that at first glance there is no variation in the colours on the map. Instead they present the binary choice, yes or no. To get the details of the vote, the user needs to select the Yes% or No%. From those we see not a lot of variation—probably not unsurprising given the overwhelmingness of the vote—as the Dublin area had the most yes, the rest of Ireland fairly solidly yes, and only Donegal in the northwest voting no, and even then, barely so.
But then we have the Guardian’s results map. And I am a wee bit lost. The bin definitions offer a bit more granular detail and so the sweeping results from the Irish Times results can here be seen as a bit more simplified. I probably would have shifted the colours and kept the yes on one side of the spectrum and not mixed the yellows and oranges into the positive, or yes, side. The stunning part of the result was, after all, that only Donegal voted no. So I would expect the colour of the choropleth to reflect that sharp break and less the gradation seen here. It’s a curious choice.
But more importantly, I am left wondering about the data, the titles, or the descriptions—I cannot be sure. The key bit is the callout of Roscommon-Galway. The text says the constituency voted 57.2% yes. But the colour would seem to indicate that it voted 65–69% no. A simple mistake? Perhaps. But then I look at the wording of the legend and maybe not. Could percentage of yes vote mean something more like the expected total or the percentage of registered voters? Probably not, but I cannot quite figure out what is going on in Roscommon-Galway. And if it is a data error, it is only made more noticeable because they point out that is one of only two constituencies described in the text.
Post script: After writing this and doing some more investigation over the long holiday weekend, I found a different map that appears to be more in sync with the results. The above was probably a mistake that just didn’t get pulled down and replaced. Below is the correct one. But it goes to show you how an incorrect graphic can cause confusion.
Credit for the Irish Times piece goes to the Irish Times graphics department.
Credit for the Guardian piece goes to the Guardian graphics department.
Monday night I was doing some work outside and when I turned around to head inside I was struck by the brilliance of an object in the night sky. I had seen the Moon rise earlier in the evening, but this was far to the east. It was identifiable as a dot, not just a speck in the night sky. As I was now intrigued I went to grab my binoculars to see if I could see Venus.
Turns out I was wrong and it was Jupiter. But then I turned my binocular-aided eyes to the west and examined the Moon. That was then I decided to try and sketch my observations, as I had done with the Eclipse.
Unfortunately, it turns out it is far more difficult to sketch in the dark then under a still semi-sunny sky. But these are my attempts to digitise those observations. And as I sat and watched, I began to notice that some faint twinkling specks near Jupiter had also moved. After I came inside, I discovered that the movement and positions hewed close to the orbits of Jupiter’s moons Ganymede and Calisto. The moving speck near the Moon I had also observed was actually the bright star Regulus. (And to be fair, it had not really moved, the Moon had moved, but I was not redrawing the Moon.)
The Moon and Regulus. The cool part is the thin ring of one of the seas that could be spotted beyond the line separating lunar day from night.
Jupiter and two of its moons. The cool thing about Jupiter is just being able to see it as a round ball in space and not a distant twinkling speck.
Here in Philadelphia, I think yesterday was the first day it had not rained in over a week. Not that everyday was a drenching storm, but at least showers passed through along with some downpours and definitely grey skies. But what about my old home, Chicago?
Well, FiveThirtyEight turned to a longer-term look and examined how over the century the amount of rainfall in the upper Midwest has been increasing. We are actually looking at the same places the Post looked at a few days ago. But instead of political maps, we have rainfall maps.
This one in particular is weird.
I get why they have the map, to show the geographic distribution of the rain gauges that collect the data. And those are site specific, not statewide. But did the designer have to choose area?
We know that area is a less than ideal way of allowing users to compare data points. And as I just noted, a choropleth, even at say the county level, is out of the question. But what about little squares? Or circles? Could colour have been used to encode the same data instead of size? And then we would likely have fewer overlapping triangles.
I suppose the argument is that the big triangles make a bigger visual impact. But they do so at the cost of comparable data points across the Midwest. Maybe the designer chose the area of triangles because there were too few gauges across the country. I am not sure, but for me the triangles are not quite on point.
That said, the graphics throughout the rest of the article are quite good, especially the opening scatterplots. They are not the sexiest of charts, but they clearly show a trends towards a wetter climate.
Back in March I posted about a great graphic from the New York Times editorial board they made in the wake of the Parkland, Florida school shooting. Saturday morning, the day after Friday’s Santa Fe, Texas school shooting, I was reading the paper and found the updated graphic.
Yeah, almost nothing has changed. Congress passed and the president signed an omnibus spending bill that included language to improve reporting on background checks.
Now from a design standpoint, what’s nice about this graphic is its restrained use of colour. The whole piece works in black and white. Of course it helps that there is nothing to show that needs to be highlighted in the data.
Credit for the piece goes to the New York Times graphics department.
It’s Friday, everybody. We made it. So now go and hit the books this weekend and study up. Thanks to xkcd, we know a little bit more about areas of research. I just am wondering where design is. Or economics.
Continuing with election-y stuff, I want to share a fascinating map from the Washington Post. The article came out last week, and it is actually incredibly light in terms of data visualisation. By my count, there were only two maps. The article’s focus is on interviews with Trump voters in 2016 and how their opinions of the president have changed over the last year or so. If you want to read it, and you should as it is very well written, I will warn you that it is long. But, to the map.
What I loved about this map is how it flips the usual narrative a bit on its head. We talk about how much a candidate won a county in 2016, or even how much the vote shifted in 2016. And anecdotally we talk about “ancestral Democrats” flipping to Trump. But this map actually tries to chart that. It reveals the last time a county actually voted for a Republican presidential candidate—the darker the red, the further back in time one has to go.
Counties that vote Democratic are white, because why do we need them for this examination. Omitting them was a great design decision. Much of the country, as we know or can intuit, voted Republican in 2012 for Mitt Romney. But what about before then? You can see how the upper Midwest, along the Mississippi River, was a stronghold for Democrats with some counties going as far back as the 1980s or earlier. And then in 2016 they all flipped and that flipping was most significant there—of some additional interest to me are the counties in Maine, the Pacific Northwest, and along Lake Erie near Cleveland.
In short, this was just a brilliantly done map. And it sets the tone for the rest of the article, which is interviews with residents of those counties called out on the map.
Credit for the piece goes to Andrew Braford, Jake Crump, Jason Bernert and Matthew Callahan.
Surprise, surprise. This morning we just take a quick little peak at some of the data visualisation from the Pennsylvania primary races yesterday. Nothing is terribly revolutionary, just well done from the Washington Post, Politico, and the New York Times.
But let’s start with my district, which was super exciting.
Each of the three I chose to highlight did a good job. The Post was very straightforward and presented each office with a toggle to separate the two parties. Usually, however, this was not terribly interesting because races like the Pennsylvania governor had one incumbent running unopposed.
But Politico was able to hand it differently and simply presented the Democratic race above the Republican and simply noted that the sitting governor ran unopposed. This differs from the Post, where it was not immediately clear that Tom Wolf, the governor, was running unopposed and had already won.
The Times handled it similarly and simultaneously displayed both parties, but kept Wolf’s race simple. The neat feature, however, was the display of select counties beneath the choropleth. This could be super helpful in the midterms in several months when key races will hinge upon particular counties.
But where the Times really shines is the race for Pennsylvania’s lieutenant governor. Fun fact, in Pennsylvania the governor and lieutenant governor do not run as a ticket and are voted for separately. This year’s Democratic incumbent, Mike Stack, does not get on with the governor and had a few little scandals to his name, prompting several Democrats to run against him. And the Times’ piece shows the two parties result, side-by-side.
Credit for the Post’s piece goes to the Washington Post graphics department.
Credit for Politico’s piece goes to Politico’s graphics department.
Credit for the Times’ piece goes to Sarah Almukhtar, Wilson Andrews, Matthew Bloch, Jeremy Bowers, Tom Giratikanon, Jasmine C. Lee and Paul Murray, and Maggie Astor.
Yesterday we talked about a static graphic from the New York Times that ran front and centre on the, well, front page. Whilst writing the piece, I recalled a piece from Politico that I have been lazily following, as in I bookmarked to write about another time. And suddenly today seemed as good as any other day.
After all, this piece also is about women running for Congress, and a bit more widely it also looks at gubernatorial races. It tracks the women candidates through the primary season. The reason I was holding off? Well, we are at the beginning of the primary season and as the Sankey diagram in the screenshot below shows, we just don’t have much data yet. And charts with “Wait, we promise we’ll have more” lack the visual impact and interest of those that are full of hundreds of data points.
But we should still look at it—and who knows, maybe late this summer or early autumn I will circle back to it. After all, today is primary day in Pennsylvania. (Note: Pennsylvania is a closed primary state, which means you have to belong to the political party to vote for its candidates.) So this tool is super useful looking ahead, because it also shows the slate of women running for positions.
I really like the piece, but as I said above, I will want to circle back to it later this year to see it with more data collected.