Correcting CBS News Charts

One of the long-running critiques of Fox News Channel’s on air graphics is that they often distort the truth. They choose questionable if not flat-out misleading baselines, scales, and adjust other elements to create differences where they don’t exist or smooth out problematic issues.

But yesterday a friend sent me a graphic that shows Fox News isn’t alone. This graphic came from CBS News and looked at the California recall election vote totals.

If you just look at the numbers, 66% and 34%, well we can see that 34 is almost half of 66. So why does the top bar look more like 2/3 of the length of the bottom? I don’t actually know the animus of the designer who created the graphic, but I hope it’s more ignorance or sloppiness than malice. I wonder if the designer simply said, 66%, well that means the top bar should be, like, two-thirds the length of the bottom.

The effect, however, makes the election seem far closer than it really was. For every yes vote, there were almost two no votes. And the above graphic does not capture that fact. And so my friend asked if I could make a graphic with the correct scale. And so I did.

One really doesn’t need a chart to compare the two numbers. And I touch on that with the last point, using two factettes to simply state the results. But let’s assume we need to make it sexy, sizzle, or flashy. Because I think every designer has heard that request.

A simple scale of 0 to 66 could work and we can see how that would differ from the original graphic. Or, if we use a scale of 0 to 100, we can see how the two bars relate to each other and to the scale of the total vote. That approach would also have allowed for a stacked bar chart as I made in the third option. The advantage there is that you can easily see the victor by who crosses the 50% line at the centre of the graphic.

Basically doing anything but what we saw in the original.

Credit for the original goes to the CBS News graphics department.

Credit for the correction is mine.

Big Beer

A few weeks back, a good friend of mine sent me this graphic from Statista that detailed the global beer industry. It showed how many of the world’s biggest brands are, in fact, owned by just a few of the biggest companies. This isn’t exactly news to either my friend or me, because we both worked in market research in our past lives, but I wanted to talk about this particular chart.

Not included, your home brew

At first glance we have a tree map, where the area of each “squarified” shape represents, usually, the share of the total. In this case, the share of global beer production in millions of hectolitres. Nothing too crazy there.

Next, colour often will represent another variable, for market share you might often see greens or blues to red that represent the recent historical growth or forecast future growth of that particular brand, company, or market. Here, however, is where the chart begins to breakdown. Colour does not appear to encode any meaningful data. It could have been used to encode data about region of origin for the parent company. Imagine blue represented European companies, red Asian, and yellow American. We would still have a similarly coloured map, sans purple and green,

But we also need to look at the data the chart communicates. We have the production in hectolitres, or the shape of the rectangle. But what about that little rectangle in the lower right corner? Is that supposed to be a different measurement or is it merely a label? Because if it’s a label, we need to compare it to the circles in the upper right. Those are labels, but they change in size whereas the rectangles change only in order to fit the number.

And what about those circles? They represent the share of total beer production. In other words the squares represent the number of hectolitres produced and the circles represent the share of hectolitres produced. Two sides of the same coin. Because we can plot this as a simple scatter plot and see that we’re really just looking at the same data.

Not the most interesting scatter plot I’ve ever seen…

We can see that there’s a pretty apparent connection between the volume of beer produced and the share of volume produced—as one would (hopefully) expect. The chart doesn’t really tell us too much other than that there are really three tiers in the Big Six of Breweries. AB Inbev is in own top tier and Heineken is a second separate tier. But Carlsberg and China Resources Snow Breweries are very competitive and then just behind them are Molson Coors and Tsingtao. But those could all be grouped into a third tier.

Another way to look at this would be to disaggregate the scatter plot into two separate bar charts.

And now to the bars…

You can see the pattern in terms of the shapes of the bars and the resulting three tiers is broadly the same. You can also see how we don’t need colour to differentiate between any of these breweries, nor does the original graphic. We could layer on additional data and information, but the original designers opted not to do that.

But I find that the big glaring miss is that the article makes the point despite the boom in craft beer in recent years, American craft beer is still a very small fraction of global beer production. The text cites a figure that isn’t included in the graphic, probably because they come from two different sources. But if we could do a bit more research we could probably fit American craft breweries into the data set and we’d get a resultant chart like this.

A better bar…

This more clearly makes the point that American craft beer is a fraction of global beer production. But it still isn’t a great chart, because it’s looking at global beer production. Instead, I would want to be able to see the share of craft brewery production in the United States.

How has that changed over the last decade? How dominant are these six big beer companies in the American market? Has that share been falling or rising? Has it been stable?

Well, I went to the original source and pulled down the data table for the Top 40 brewers. I took the Top 15 in beer production, all above 1% share in 2020, and then plotted that against the change in their beer production from 2019 to 2020. I added a benchmark of global beer production—down nearly 5% in the pandemic year—and then coloured the dots by the region of origin. (San Miguel might not seem to fit in Asia by name, but it’s from the Philippines.)

Now I can use a good bar.

What mine does not do, because I couldn’t find a good (and convenient) source is what top brands belong to which parent companies. That’s probably buried in a report somewhere. But whilst market share data and analysis used to be my job, as I alluded to in the opening, it is no longer and I’ve got to get (virtually) to my day job.

Credit to the original goes to Felix Richter.

Credit for my take goes to me.

If You Can’t Stand the Heat, Cut Your Carbon Emissions Pt. II

A few weeks ago I wrote about the United Nation’s Intergovernmental Panel on Climate Change (IPCC) latest report on climate change, which synthesised the last several years’ data. If you didn’t see that post, suffice it to say things are bad and getting worse. At the time I said I wanted to return to talk about a few more graphics in the release. Well, here we are.

In this piece we have a map, three technically. In a set of small multiples, the report’s designers show the observed change, i.e. what’s happening today, and the degree of scientific consensus on whether humans are causing it.

It’s gotten hotter and wetter here in eastern North America

What I like about this is that, first, improved data and accuracy allows for sub-continental breakdowns of climate change’s impacts. That breakdown allowed the designers to use a tilemap consisting of hexagons to map those changes.

Since we don’t look at the world in this kind of way, the page also includes a generous note where it defines all these acronyms. Of course even with those, it still doesn’t look super accurate—and that is fine, because that’s the point—so little strokes outside clusters of hexagons are labelled to further help the reader identify the geographic regions. I really like this part.

I also like how little dots represent the degree of confidence. The hexagons give enough space to include dots and labels while still allowing the colours to shine. These are really nice.

But then we get to colour, the one part of this graphic with which I’m not totally thrilled. The first map looks at temperature, specifically heat extremes. Red means increase in heat extremes and blue means decrease. Fair enough. Hatched pattern means there is low consensus and medium grey means there’s little data. I like it.

Moving to the second map we look at heavy precipitation. Green means an increase and yellow a decrease. Hatched and medium grey both mean the same as before. I like this too. Sure, with clear titling you could still use the same colours as the first map, but I’ll buy if you’re selling you want visual distinction from the red–blue map above.

Then we get to the third map and now we’re looking at drought. Hatched and grey mean the same. Good. But now we have green and yellow, the same green and yellow as the second map. Okay…but I thought the second map showed we need a visual distinction from the first? But what makes it really difficult is that in this third map we invert the meaning of green and yellow. Green now means a decrease in drought and yellow an increase.

I can get that a decrease in drought means green fields and an increase in drought means dead and dying fields, yellow or brown. And sure, red and blue relate to hot and cold. But the problem is that we have the exact same colours meaning the opposite things when it comes to precipitation.

Why not use two other colours for precipitation? You wouldn’t want to use blue, because you’re using blue in the first map. But what about purple and orange, like I often do here on Coffeespoons? This is why I don’t think the designers needed to switch up the colours from map to map. Pick a less relational colour palette, say purple and orange, and colour all three maps with purple being an increase and orange being a decrease.

Colour is my big knock on these graphics, which unfortunately could otherwise have been particularly strong. Of course, I can’t blame designers for going with red and blue for hot and cold temperatures. I’ve had the same request in my career. But it doesn’t make reading these charts any easier.

Credit for the piece goes to the IPCC graphics team.

Out with the New, In with the Old

After twenty years out of power, the Taliban in Afghanistan are back in power as the Afghan government collapsed spectacularly this past weekend. In most provinces and districts, government forces surrendered without firing a shot. And if you’re going to beat an army in the field, you generally need to, you know, fight if you expect to beat them.

I held off on posting anything about the Taliban takeover of Afghanistan simply because it happened so quick. It was not even two months ago when they began their offensive. But whenever I started to prepare a post, things would be drastically different by the next morning.

And so this timeline graphic from the BBC does a good job of capturing the rapid collapse of the Afghan state. It starts in early July with a mixture of blue, orange, and red—we’ll come back to the colours a bit later. Blue represents the Afghan government, red the Taliban, and orange contested areas.

The start of the summer offensive

The graphic includes some controls at the bottom, a play/pause and forward/backward skip buttons. The geographic units are districts, sub-provincial level units that I would imagine are roughly analogous to US counties, but that’s supposition on my part. Additionally the map includes little markers for some of the country’s key cities. Finally in the lower right we have a little scorecard of sorts, showing how many of the nearly 400 districts were in the control of which group.

Skip forward five weeks and the situation could not be more different.

So much for 20 years.

Almost all of Afghanistan is under the control of the Taliban. There’s not a whole lot else to say about that fact. The army largely surrendered without firing a shot. Though some special forces and commando units held out under siege, notably in Kandahar where a commando unit held the airport until after the government fell only to be evacuated to the still-US-held Hamid Karzai International Airport in Kabul.

My personal thoughts, well you can blame Biden and the US for a rushed US exodus that looks bad optically, but the American withdrawal plan, initiated by Trump let’s not forget, counted on the Afghan army actually fighting the Taliban and the government negotiating some kind of settlement with the Taliban. Neither happened. And so the end came far quicker than anyone thought possible.

But we’re here to talk graphics.

In general I like this. I prefer this district-level map to some of the similar province-level maps I have seen, because this gives a more granular view of the situation on the ground. Ideally I would have included a thicker line weight to denote the provinces, but again if it’s one or the other I’d opt for district-level data.

That said, I’d probably have used white lines instead of black. If you look in the east, especially south and east of Kabul, the geographically small areas begin to clump up into a mass of shapes made dark by the black outlines. That black is, of course, darker than the reds, blues, and yellows. If the designers had opted for white or even a light shade of grey, we would enhance the user’s ability to see the district-level data by dropping the borders to the back of the visual hierarchy.

Finally with colours, I’m not sure I understand the rationale behind the red, blue, yellow here. Let’s compare the BBC’s colour choice to that of the Economist. (Initially I was going to focus on the Economist’s graphics, but last minute change of plans.)

Another day, more losses for the government

Here we see a similar scheme: red for the Taliban, blue for the government. But notably the designers coloured the contested areas grey, not yellow. We also have more desaturated colours here, not the bright and vibrant reds, blues, and yellows of the BBC maps above.

First the grey vs. yellow. It depends on what the designers wanted to show. The grey moves the contested districts into the background, focusing the reader’s attention on the (dwindling) number of districts under government control. If the goal is to show where the fighting is occurring, i.e. the contest, the yellow works well as it draws the reader’s attention. But if the goal is to show which parts of the country the Taliban control and which parts the government, the grey works better. It’s a subtle difference, I know, but that’s why it would be important to know the designer’s goal.

I’ll also add that the Economist map here shows the provincial capitals and uses a darker, more saturated red dot to indicate if they’d fallen to the Taliban. Contrast that with the BBC’s simple black dots. We had a subtler story than “Taliban overruns country” in Afghanistan where the Taliban largely did hold the rural, lower populated districts outside the major cities, but that the cities like the aforementioned Kandahar, Herat, Mazar-i-Sharif held out a little bit longer, usually behind commando units or local militia. Personally I would have added a darker, more saturated blue dot for cities like Kabul, which at the time of the Economist’s map, was not under threat.

Then we have the saturation element of the red and blue.

Should the reds be brighter, vibrant and attention grabbing or ought they be lighter and restrained, more muted? It’s actually a fairly complex answer and the answer is ultimately “it depends”. I know that’s the cheap way out, but let me explain in the context of these maps.

Choropleth maps like this, i.e. maps where a geographical unit is coloured based on some kind of data point, in this instance political/military control, are, broadly speaking, comprised of large shapes or blocks of colour. In other words, they are not dot plots or line charts where we have small or thin instances of colour.

Now, I’m certain that in the past you’ve seen a wall or a painting or an advert for something where the artist or designer used a large, vast area of a bright colour, so bright that it hurt your eyes to look at the area. I mean imagine if the walls in your room were painted that bright yellow colour of warning signs or taxis.

That same concept also applies to maps, data visualisation, and design. We use bright colours to draw attention, but ideally do so sparingly. Larger areas or fields of colours often warrant more muted colours, leaving any bright uses to highlight particular areas of attention or concern.

Imagine that the designers wanted to highlight a particular district in the maps above. The Economist’s map is better designed to handle that need, a district could have its red turned to 11, so to speak, to visually separate it from the other red districts. But with the BBC map, that option is largely off the table because the colours are already at 11.

Why do we have bright colours? Well over the years I’ve heard a number of reasons. Clients ask for graphics to be “exciting”, “flashy”, “make it sizzle” because colours like the Economist’s are “boring”, “not sexy”.

The point of good data visualisation, however, is not to make things sexy, exciting, or flashy. Rather the goal is clear communication. And a more restrained palette leaves more options for further clarification. The architect Mies van der Rohe famously said “less is more”. Just as there are different styles of architecture we have different styles of design. And personally my style is of the more restrained variety. Using less leaves room for more.

Note how the Economist’s map is able to layer labels and annotations atop the map. The more muted and desaturated reds, blues, and greys also allow for text and other artwork to layer atop the map but, crucially, still be legible. Imagine trying to read the same sorts of labels on the BBC map. It’s difficult to do, and you know that it is because the BBC designers needed to move the city labels off the map itself in order to make them legible.

Both sets of maps are strong in their own right. But the ultimate loser here is going to be the Afghan people. Though it is pretty clear that this was the ultimate result. There just wasn’t enough support in the broader country to prop up a Western style liberal democracy. Or else somebody would have fought for it.

Credit for the BBC piece goes to the BBC graphics department.

Credit for the Economist piece goes to the Economist graphics department.

If You Can’t Stand the Heat, Cut Your Carbon Emissions

Earlier this morning (East Coast time) the Intergovernmental Panel on Climate Change (IPCC), the UN’s committee studying climate change, released its latest review of climate change. This is the first major review since 2013 and, spoiler, it’s not good.

I’ve read some news articles about the findings, but I want to critique and comment upon some of the graphics contained within the report itself. This started going too long, however, so I think I will break this into several shorter, more digestible chunks.

And I want to start with the first chart, two line charts that lay out the temperature changes we’ve seen.

Going up…

One of the first things I like here is the language. Often we might see these or similar charts that simply state temperatures from the year 1 through 2020. One of the common reasons I hear from people that deny climate change is that “people weren’t recording temperatures back in 1 AD.

They would be correct. We do not have planet-wide meteorological observations from the time of Julius Caesar. But in the year 2021 we do have science. And that allows us to take other evidence, e.g. dissolved carbon dioxide in ice, or tree ring size, &c., and use them to reconstruct the temperature record indirectly.

And reconstruct is the word the IPCC uses to clearly delineate the temperature data pre- and post-1850 when their observed data set begins.

The designers then highlight this observed data set, broadly coinciding with the Industrial Revolution when we as a species began to first emit extra greenhouse gasses into the atmosphere. You can see this as a faint grey background and a darker stroke along the x-axis.

Additionally, the designers used annotations to call out the first main point, that warming in the last almost two centuries is far beyond what we’ve seen in the last two millennia.

The second annotation points to a bar, reminiscent of the range of a box plot, that exists outside the x-axis and almost embedded within the y-axis. This bar captures the range of temperatures reconstructed in the past 100,000 years. And by including it in the chart, we can see that we have just recently begun to exceed even that range.

In the second chart, we have the entire background shaded light grey and the whole x-axis in a darker stroke to remind us that we are now looking at the Industrial/Post-Industrial era. But what this chart does is do what scientists do, test whether natural, non-manmade causes can fully explain the temperature increase.

They can’t.

The chart plots the modelled data looking at just natural causes vs. modelled data looking at natural causes plus human impacts. Those lines and their ranges are then compared to the temperatures we’ve observed and recorded.

Since the 1930s and 40s, it’s been a pretty clear and consistent tracking with natural plus manmade causes. For years the scientific community has been in agreement that humanity is contributing to the rising temperatures. This is yet more evidence to make the point even more conclusively.

These are two really good charts that taken together show pretty conclusively that humanity is directly responsible for a significant portion of Earth’s recent climate change.

I’ll have more on some other notable graphics in the report later in the week, so stay tuned.

Credit for the piece goes to the IPCC graphics team.

Inflating Areas

One trend people have begun to follow lately is that of rising prices for consumer goods. If you have shopped recently for things, you may have noticed that you have been paying more than you were just a few weeks ago. We call this inflation. The Bureau of Labour Statistics (BLS) tracks this for a whole range of goods. We call the the consumer price index (CPI)

Prices can vary wildly for some goods, most notably food and energy. For those of my readers who drive, recall how quickly petrol/gasoline prices can change. Because of that volatility, the Bureau of Labour Statistics strips out food and energy prices and the inflation that excludes food and energy is what we call Core CPI.

Lately, we have been seeing an increase in prices and inflation is on the rise. To an extent, this is not surprising. The pandemic disrupted supply chains and wiped out supplies and stores of goods. But with many people working remotely, many now have pent up savings they want to spend. But with low supply and high demand, basic economics suggests rising prices. As supplies increase in the coming months, however, the rise in prices will begin to cool off. In other words, most economists are not yet concerned and expect this spike in inflation to be passing in nature. But not everyone agrees.

Last week, the Washington Post had an article examining the cause of inflation for a number of industries. To do so, it used some charts looking at prices over the past two years. This screenshot is from the used car section.

Going up…

I want to focus on the design of this graphic, though, not the content. The designers’ goal appears to be contrasting the inflation over the last year to that of the last two years. Easy peasy. Red represents one-year inflation and blue two-year.

Typically when you see a chart that look like this, an area or filled line chart, the coloured area reflects the total value of the thing being measured. You can also use the colour to make positive/negative values clearer. In this case, neither of those things are happening.

Because the blue, for example, starts at the beginning of the time series and at the bottom of the chart, it looks like an enormous amount of consistent blue growth. And when the line runs into May 2020, we begin to see what appears as a stacked area chart, with the blue area increasing at the expense of the red.

Another way of reading it could be that the 29.7% and 29.3% increases equal the shaded areas, but that’s also problematic. If the shaded area locked to the baseline like you’ll see in a moment, I could maybe see that working, but at this point it just leaves me confused.

Now you can use the area fill to make it clear when a line dips above or below the baseline, in this case 0%. And I took that approach when I reimagined the chart as seen below.

The earlier chart, reimagined

What we do here is we set the bottom of the area fill to the baseline. Consequently, where the chart is filled above 0 we have positive inflation, and where it falls below the 0 line we have negative inflation, or deflation.

We need to note here that the text in the original article talks about the monthly change in inflation, e.g. that used car prices have increased by 7.3% last month. That, however, is not what the chart looks at. Instead, the chart shows the change yearly, in other words, prices now vs last May. To an extent, the 29.7% increase is not terribly surprising given how terrible the recession was.

Ultimately, I don’t see the value in the filled blue and red areas of the chart because I am left more confused. Does the reader need to see how far back one year and two years are from May 2021? Don’t the date labels do that sufficiently well?

This is just a weird article that left me scratching my head at the graphics. But read the text, it’s super informative about the content. I just wish a bit more work went into the graphics. There are some nice illustrations beginning each section, but I kind of feel that more time was spent on the illustrations than the charts.

Credit for the piece goes to Abha Bhattarai and Alyssa Fowers.

On a Line. Or Not.

Two weeks ago I was reading an article in the BBC that fact checked some of President Biden’s claims about the economy. Now I noted the other day in a post about axis lines and their use in graphics. Axis lines help ground the user in making comparisons between bars, lines, or whatever, and the minimum/maximum/intervals of the data set.

I was reading the article and first came upon this graphic. It’s nothing crazy and shows job growth in the aggregate for the first three months of a presidential administration. A pretty neat comparison in the combination of the data. I like.

Pay attention to what you see here. There will be a quiz.

I don’t like the lack of grid lines for the axis, however. But, okay, none to be found.

I keep reading the article. And then a couple of paragraphs later I come upon this graphic. It looks at the monthly figures and uses a benchmark line, the red dotted one, to break out those after January 2021 when Biden took office.

Spot the differences.

But do you notice anything?

The lines for the y-axis are back!

The article had a third graphic that also included axis lines.

I don’t have a lot to say about these graphics in particular, but the most important thing is to try and be consistent. I understand the need to experiment with styles as a brand evolves. Swap out the colours, change the styles of the lines, try a new typeface. (Except for the blue, we are seeing different colours and typefaces here, but that’s not what I want to write about.)

First, I don’t know if these are necessarily style experiments. I suspect not, but let’s be charitable for the sake of argument. I would refrain from experimenting within a single article. In other words, use the lines or don’t, but be consistent within the article.

For the record, I think they should use the lines.

Another point I want to make is with the third graphic. You’ll note that, like I said above, it does use axis lines. But that’s not what I want to mention.

At least we have lines.

Instead I want to look at the labelling on the axes. Let’s start with the y-axis, the percentage change in GDP on the previous quarter. The top of the chart we have 30%. As I’ve said before, you can see in the Trump administration, the bar for the initial Covid-19 rebound rises above the 30% line. It’s not excessive, I can buy it if you’re selling it.

But let’s go down below the 0-line. Just prior to the rebound we had the crash. Similarly, this extends just below the -30% line. But here we have a big space and then a heavy black line below that -30% line. It looks like the bottom line should be -40%, but scanning over to the left and there is no label. So what’s going on?

First, that heavy black line, why does it appear the same as the baseline or zero-growth line? The axis lines, by comparison, are thin and grey. You use a heavier, darker line to signify the breaking point or division between, in this case, positive and negative growth. Theoretically, you don’t need the two different colours for positive and negative growth, because the direction of the bar above/below that black line encodes that value. By making the bottom line the same style as the baseline, you conflate the meaning of the two lines, especially since there is no labelling for the bottom line to tell you what the line means.

Second, the heaviness of the line draws visual attention to it and away from the baseline, especially since the bottom line has the white space above it from the -30% line. Consider here the necessity of this line. For the 30% line that sets the maximum value of the y-axis, we have the blue bar rising above the line and the administration labels sit nicely above that line. There is no reason the x-axis labels could not exist in a similar fashion below the -30% line. If anything, this is an inconsistency within the one chart, let alone the one graphic.

Third, is it -40%? I contend the line isn’t necessary and that if the blue bar pokes above the 30% line, the orange bar should poke below the -30% line. But, if the designer wants to use a line below the -30% line, it should be labelled.

Finally, look at the x-axis. This is more of a minor quibble, but while we’re here…. Look at the intervals of the years. 2012, 2014, 2016, every two years. Good, make sense. 2018. 20…21? Suddenly we jump from every two years to a three-year interval. I understand it to a point, after all, who doesn’t want to forget 2020. But in all seriousness, the chart ends at 2021 and you cannot divide that evenly. So what is a designer to do? If this chart had less space on the x-axis and the years were more compressed in terms of their spacing, I probably wouldn’t bring this up.

However, we have space here. If we kept to a two-year interval system, I would introduce the labels as 2012, but then contract them with an apostrophe after that point. For example, 2014 becomes ’14. By doing that, you should be able to fit the two-year intervals in the space as well as the ending year of the data set.

Overall, I have to say that this piece shocked me. The lack of attention to detail, the inconsistency, the clumsiness of the design and presentation. I would expect this from a lesser oganisation than the BBC, which for years had been doing solid, quality work.

The first chart is conceptually solid. If Biden spoke about job creation in the first three months of the administration vs. his predecessor, aggregate the data and show it that way. But the presentation throughout this piece does that story a disservice. I wish I knew what was going on.

Credit for the piece goes to the BBC graphics department.

The May Jobs Report

Last Friday, the government released the labour statistics from April and they showed a weaker rebound in employment than many had forecasted. When I opened the door Saturday morning, I got to see the numbers above the fold on the front page of the New York Times.

Welcome to the weekend

What I enjoyed about this layout, was that the graphic occupied half the above the fold space. But, because the designers laid the page out using a six-column grid, we can see just how they did it. Because this graphic is itself laid out in the column widths of the page itself. That allows the leftmost column of the page to run an unrelated story whilst the jobs numbers occupy 5/6 of the page’s columns.

If we look at the graphic in more detail, the designers made a few interesting decisions here.

Jobs in detail

First, last week I discussed a piece from the Times wherein they did not use axis labels to ground the dataset for the reader. Here we have axis labels back, and the reader can judge where intervening data points fall between the two. For attention to detail, note that under Retail, Education and health, and Business and professional services, the “illion” in -2 Million was removed so as not to interfere with legibility of the graphic, because of bars being otherwise in the way.

My issue with the axis labels? I have mentioned in the past that I don’t think a designer always needs to put the maximum axis line in place, especially when the data point darts just above or below the line. We see this often here, for example Construction and Manufacturing both handle it this way for their minimums. This works for me.

But for the column above Construction, i.e. State and local government and Education and health, we enter the space where I think the graphic needs those axis lines. For Education and health, it’s pretty simple, the red losses column looks much closer to a -3 million value than a -2 million value. But how close? We cannot tell with an axis line.

And then under State and local government we have the trickier issue. But I think that’s also precisely why this could use some axis lines. First, almost all the columns fall below the -1 million line. This isn’t the case of just one or two columns, it’s all but two of them. Second, these columns are all fairly well down below the -1 million axis line. These aren’t just a bit over, most are somewhere between half to two-thirds beyond. But they are also not quite nearly as far to -2 million as the ones we had in the Education and health growth were near to -3 million.

So why would I opt to have an axis line for State and local governments? The designers chose this group to add the legend “Gain in April”. That could neatly tuck into the space between the columns and the axis line.

Overall it’s a solid piece, but it needs a few tweaks to improve its legibility and take it over the line.

Credit for the piece goes to Ella Koeze and Bill Marsh.

Off the Axis

Two Fridays ago, I opened the door and found my copy of the New York Times with a nice graphic above the fold. This followed the announcement from the White House of aggressive targets to reduce greenhouse gas emissions

In general, I love seeing charts and graphics above the fold. As an added bonus, this set looked at climate data.

Need to see more downward trending lines.

But there are a few things worth pointing out.

First from a data side, this chart is a little misleading. Without a doubt, carbon dioxide represents the greatest share of greenhouse gasses, according to the US Environmental Protection Agency (EPA) it was 76% in 2010. Methane contributes the next largest share at 16%. But the labelling should be a little clearer here. Or, perhaps lead with a small chart showing CO2’s share of greenhouse gasses and from there, take a look at the largest CO2 emitters per person.

Second, where are the axis labels?

I will probably have more on this at a later date, but neither the bar chart nor the line charts have axis labels. Now the designers did choose to label the beginning value for the lines and the bars, but this does not account for the minimums or maximums. (It also assumes that the bottom of the lines is zero.)

For example, we can see that China began 1990 with emissions at 3.4 billon metric tons. The annotation makes clear that China’s aggregate emissions surpassed those of the US in 2004. But where do they peak? What about developing countries?

If I pull out a ruler and draw some lines I can roughly make some height comparisons. But, an easier way would be simply to throw some dotted lines across the width of the page, or each line chart.

This piece takes a big swing at presenting the challenge of reducing emissions, but it fails to provide the reader with the proper—and I think necessary—context.

Credit for the piece goes to Nadja Popovich and Bill Marsh.

Arrowheads

I don’t know if this is a trend, but I’ve now seen a few graphics appearing using arrows to show the direction or trend of the data. This graphic in an article by Bloomberg prompted me to talk about this piece.

I should add, after rereading my draft, that I’m not clear who made this graphic. I assume that it was the Bloomberg graphics team, because it appears in Bloomberg and all the data is presented to recreate the chart. But, it could also be a chart made by someone at Goldman Sachs that credits Bloomberg as a source and then someone at Bloomberg got hold of a copy. And a graphic made for a news/media outlet will typically be of a different quality or level of polish than one made perhaps by and for analysts. (Not that I think there should be said differences, as it does a disservice to internal users, but I digress from a digression.)

All the things going on in this chart.

The arrow here appears above the peak quarter, i.e. the second of 2021, for both the Goldman Sachs Economics forecast and the consensus forecast. But what does it really add? First, it adds “ink”, in this case pixels. Here, every pixel consumes our attention and there is a finite number of available pixels within the space of this graphic.

When I work with authors or subject matter experts, I often find myself asking them “what’s the most important thing to communicate?” or something along those lines. If the person answers with a long laundry list, I remind them that if everything is important, nothing is important. If everything is set in bold, all caps text, what will look most important is the rare bit of text set in regular, lower-case letters.

In the above graphic, there are so many things screaming for my attention, it’s difficult to say which is the most important. First, I’m fairly certain that “US QoQ annualised GDP growth” could move to the graphic subhead or data definition. Allow the graphic’s data container to contain, well, data. Second, the data series labels can be moved outside the data container. The labels here have an inherent problem is that the Goldman Sachs Economics numbers are in blue, and that blue text has less visual weight than the black text of the Consensus label. Consequently, the Goldman Sachs Economics label recedes into the background and becomes lost, not what you want from your legend.

Third, I don’t believe the data labels here add anything to the chart. They function as sparkly distractions from the visual trend, which should be the most important aspect of a visual chart.

Finally, we get to the arrow, the impetus for this post. First, I should note that it is not clear what growth it shows. The fact the line is black makes me think it reflects the Consensus forecast whereas a blue line would represent the Goldman Sachs forecast. But it could also be the average of the two or even a more general “here’s the general shape”. The problem is that the shape matters. If you look at the slope of the actual forecasts, you see a sharp increase to the peak followed by a slower, more gradual taper. The arrow in the original graphic shows a decelerating curve that is shallower in the lead up to the peak and that is not what is forecast to happen.

Now we get to the issue I mentioned at the top, the extraneous labelling and data ink wasted. If we look at the chart as is, but remove the arrow, we see this.

Immediately to the right of the peak, we have have some blue data labels and then just a bit to the right of that, but sitting vertically above the label we have the bold blue text labelling the data series. But further to the upper right we have a dark and bold block of text that draws the eye away from the peak and into the corner. It draws the eye away from the very element of the shape the peak needs to be a peak, the trough in the wave. Consequently, it makes sense with the eye being drawn up and to the right that the designers threw an arrow in above the peak to show how, no, actually your eye needs to go down and to the right.

But what happens if we then strip out the data series labelling? Do we still need the arrow? Let’s take a look.

I would argue that no, we do not. And so let’s strip the arrow out of the picture and take a look.

Here the shape of the curve is clear, a sharp rise and then a gradual taper to the right. No arrow needed to show the contour. In other words, the additional labelling wastes our attention, which then forces us to add an arrow to see what we needed to see in the first place, but then further wasting our attention.

There are a number of other things I take issue with in this chart: the black outlines of the blue rectangles, the tick marks on the x-axis, the solid border of the container, the lack of axis lines. But the arrow points to this graphic’s central problem, a poorly thought out labelling structure.

So because the chart provides all the data, I took a quick stab at how I would chart it using my own styles. I gave myself a 3:2 ratio, less space than the original graphic had. This is where I landed. I would prefer the legend below the chart labelling, but it felt cramped in the space. And with so few data points along the x-axis, the chart doesn’t need a ton of horizontal space and so I repurposed some of it to create a vertical legend space.

I mixed typefaces only because my default does not have a proper small capitals and I wanted to use small capitals to reduce and balance out the weight of the exhibit label in the graphic title.

I could still tweak the spacing between the bars and perhaps the treatment of the years below the quarters could use some additional work, but the main point here is that the shape of the curve is clear. I need no arrow to tell the user that there is a peak and that after the peak the line goes down. The white space around the bars and the line does that for me.

Credit for the piece goes to either the Bloomberg graphics department or the Goldman Sachs graphics department. Not sure.