Graphic Sociology

American Generation Age Timeline (Age measured in 2009) | Pew Research

What works

Pew Research has created a tidy series of interactive graphics to describe the demographic characteristics of American generational cohorts from the the Silent Generation (born 1928 – 1945) through the Boomers (born 1946 – 1964), Generation X (1965 – 1980), and the Millennial Generation (born 1981+). The interactive graphics frame the data well. They offer the timeline above as contextual background and a graphic way to describe what they .

Then users can flip back and forth between comparing each generation to another along a range of variables – labor force participation, education, household income, marital status – while they were in the 18-29 year old age group OR by looking at where each generation is now. The ability to interact makes the presentation extremely illustrative and pedagogically meaningful. It is much easier to understand patterns that are changing over time versus patterns that are life course specific.

Marital status

Marital status by generation measured when young | Pew Research

Marital status by generation measured in 2009 (snapshot) | Pew Research

For instance, marital trends have been hard to talk about because the age at first marriage moves up over time, so it’s hard to figure out at what age we can expect that people will have gotten married if they are ever going to do so (I tried looking at marriage here).

What I like about the Pew Research graphics is that they show us not only what the generations looked like when they were between 18 and 29 years old (above) and again what they look like now (below). Not only does it become obvious how many more people are choosing to remain unmarried, but it also becomes clear that in addition to divorce, widowhood is a major contributor to the end of marriage.

Labor force participation

Here’s another set of “when they were young” vs. “where they are now” comparisons, this time on labor force participation. It appears that the recession has walloped the youngest, least experienced workers the hardest. They have the highest unemployment rate AND the highest rate of educational attainment (and school loan debt), which leaves them much worse off as they start out than their parents were in the Boomer Generation. Even if their parents were in Generation X, they were still better off than today’s 20-something Millennials.

American labor force participation by generation (2009) | Pew Research

American Labor Force Participation by Generation (measured in 2009) | Pew Research

What needs work – Are generations meaningful?

My first minor complaint is that the graphic does not make clear *exactly* what “when they were young” means. If we look at the first graphic in the series, the timeline, it appears that “when they were young” was measured when each generation was between 18 and 29 years old. I hope that is the case. I might have had an asterisk somewhere explaining that “when they were young = when they were 18-29 years old”.

The concept of generations, in my opinion, is a head-scratcher. What is the analytical purchase of looking at generations – strictly birth-year delimited groups that supposedly share a greater internal coherence than other affinal or ascribed statuses we might imagine? If we believe that social, technological, and most all kinds of change happen over time, of course there are going to be measurable differences between one generation and the next. I imagine, though I have never seen the comparison, that if social scientists split people into 10- or 20-year pools based on their birth years they would end up with the same sorts of results. So why not think of generations as even units? And is it clear that the meaningful changes are happening in 20-year cycles? Or would 10-year age cohorts also work?

The real trickiness comes in when we think about individuals. Say someone is like myself, born in a year on the border between one generation and the next. Am I going to be just as much like a person born firmly in the middle of my cohort as a person on the far end of it? Or will people like me have about as much in common with the people about 8 years above and below us, but less in common with the people 15 years older than us who are considered to be in the same generation, and thus to have many similar tendencies/life chances/characteristics?

A better way to measure the cohort effect would seem to be to consider each individual’s age distance from each other individual in the sample – the closer we are in age, the more similar we could be expected to be with respect to things like labor force participation and educational attainment. Large structural realities like recessions are going to hit us all when we have roughly similar amounts of work force experience, impacting us similarly (though someone 10 years older and still officially in the same generation will probably fare much better). Since it is computationally possible to run models that can take the actual age distances of individuals in the same into account, I don’t understand the analytical purchase of the concept of generations.

The take-away: great graphics, bad premise.

References

Taylor, Paul and Keeter, Scott, eds. (24 February 2011) The Millenials. Confident. Connected. Open to Change. [Full Report] [See also: Executive Summary and Interactive Infographic] Washington, DC: Pew Research Center.

Reading, Writing, and Earning Money | GOOD Transparency Blog

What works

Nothing is working for me with this graphic except possibly the few places where the designers offered detailed information about a particular location’s high school graduation ranking, college graduation ranking, and income ranking. But that’s being generous.

What needs work

Horrible use of a map. Maps should only be used where there is good reason to believe the information being conveyed is tied closely to geography. This information is not tied closely to geography though it might be tied closely to states. But states need not always be represented as geographical entities. Often, they are political entities and their particular geography is not salient.

The math that led to the graphic flattens important details and renders this a useless graphic. What I believe the designers did was something like this:

• They took all of their numbers and turned them into some scale between 0 and 100%
• Then they decided to represent each of the three variables with pure Cyan, Magenta, or Yellow. The higher the state scored on the scale from 0-100, the more saturated the color value.
• Then they gave each county a combined score by building new colors from mixing the values of the previous three. Higher scoring states ended up with more saturated colors. Basically, higher scoring states started to approach black. States that scored high on just one vector ended up having a clearer, lighter color profile.

Here’s the big problem with this. It was hard for me to explain to my MIT-educated friend so I’m not sure this is going to make sense the first time ’round. Representing everything on a scale from 0-100 is a slide towards obfuscation. The graduation rates are both unadulterated rates. The income data represents un-scaled median incomes. I appreciate that they are not scaled, but I have a hard time adding 65% with $45,000. That’s some troubled math. At least in the monochrome maps we know what we’re looking at before the three variables get added up.

A grave sin was committed when the numbers for these three different variables were added up. Now, of course, it wasn’t the numbers that were added up. It was the color values of each of the three separate data points that were added up. Additive color seems to be something that does not send up a red flag. I can guarantee you that if they had presented something – a table or graph – where they had ended up adding values from high school graduation, college graduation, and income, red flags would have been flying. Why? Well, maybe you’re starting to catch my drift, but I’ll help you by spelling it out. What happens when the colors are added is a clear violation of the ‘apples to apples’ rule. Comparisons do not work unless you are sure you are comparing like things. Graduation rates are not like income. They are two different kinds of numbers – one is a rate the other is either a linear value or a log-linear value. Either way, they cannot be added up and still make sense. It’s no surprise that the graphic ends up looking like an incomprehensible slurry of a gray area.

References

GOOD and Gregory Hubacek. (March 2011) Reading, Writing, and Earning Money in GOOD Transparency Blog.

A recent (well, 2010 so not *that* recent) report from the UNDP traces the history of information graphics as tools for the promotion of public health. Illustrious crusaders from the yesteryear of public health like John Snow and Florence Nightingale developed some of the earliest ‘infographics’ in service of their public health goals. I’ll post more on that portion of the report later this week. But for now, I’d like to discuss the bulk of the report which was dedicated to the decisions that César Hidalgo (Professor at MIT’s Media Lab, Student at Harvard’s Center for International Development, Associate Professor at Northeastern’s School of Art and Design) made as he developed an appropriate information graphic to represent country level data generated by the Human Development Index. (See also: Measure of America’s Human Development Index graphics for the US only and the Graphic Sociology post about them).

The graphic below this is not intended to be a graphic. It is the basic formula upon which the Human Development Index (HDI) is based. The HDI is a single number that represents a composite score that takes contributions from educational, income, and health measures (which are themselves composite scores). The authors first came up with a simple, almost graphical representation of the relationship between the contributing factors that’s a sort of formula/graphic hybrid. Many social scientists would stop here and move on to the writing of the report, content to let a table with country-level data do the reporting for them.

Basic Human Development Index Relationship | César Hidalgo

HDI Spline Tree

From this hybrid between a formula and a graphic, Hidalgo developed the spline tree you see below. It shares some aspects with the basic formula above, that much is visually clear, but already the lengths and colors of the components are taking on meaning, allowing each country/year combination to produce a tree that is distinct from other trees, but similar enough to be comparable.

The HDI Tree - Spline Design | César Hidalgo

One of designer’s common strengths/weaknesses is the inability to stop designing. Design is never done because the design has reached some obvious and agreed-upon level of perfection. Usually design is deemed ‘done’ when the deadline rolls around. It would appear that Hidalgo was ahead of schedule and decided to go for another iteration, coming up with the diamond tree you see below. Though, as you’ll also see, he did not completely abandon the spline tree. It shows up again.

HDI Diamond Tree

The diamond HDI tree takes an area-based approach, one that is easier to understand visually at first glance than the spline tree. With the spline tree approach, the challenge is that the viewer needs to visually compare the lengths of lines that are not parallel to one another to gain full comprehension. Granted, one might most often compare the lengths of lines that ARE parallel to one another because viewers might mostly be comparing one country to the next. But that isn’t always the case. And even that is not as easy as comparing the areas in the diamond tree approach.

The Human Development Tree - Diamond Tree | César Hidalgo

The rules for the HDI diamond tree (and I’m quoting Hidalgo and team here) are as follows:

* The height of the tree trunk is proportional to the total value of the HDI
* The side of the tree branches are proportional to each sub-indicator
* The branches are ordered in increasing order from left to right
* The color of the trunk is the average color of the components

All together

And here is one country’s worth of Diamond and Spline trees, represented over time. This is where I think the two tree graphics – and the diamond tree in particular – work their magic best. Human eyes are good at doing comparison’s in this sort of way. The trees are more or less the same thing over and over again so this repetitive presentation allows the eye to pick out the relatively small changes over time, especially as they aggregate from one year to the next.

HDI in Rwanda 1970-2005 | César Hidalgo

Pan-Africa

With the last graphic in the series, you can see what it would look like to present the entire continent of Africa, by country, in two different years. It’s a little tough to fit a properly sized graphic into the format of the blog. I encourage you to click through to the full report in the references where you can see a much better version of the final graphic.

Human Development in Africa by country, 1970-2005 | César Hidalgo

Kudos

My biggest applause goes out to the Hidalgo team for abandoning the use of any map at all. This graphic should prove the point that just because one is faced with country level data – something that seems geographical in nature – one should not feel that they must use a map. A map would not have added anything to this information and it probably would have precluded the development of the tree concepts that are working pretty well.

References

Hidalgo, César A. (2010) Graphical Statistical Methods for the Representation of the Human Development Index and its Components [Research Paper] United Nations Development Program.

In theory there is an interactive portal for comparing any two HDI Diamond Trees of your choosing but I was not able to get it to work in Firefox. Worked like a charm in Safari and Chrome.

For ongoing comments on these graphics see: The HDI Tree: A visual representation at “Let’s Talk Human Development” a website published by the United Nations Development Project.

American Shame | Charles Blow for the New York Times

What works

To social scientists: you can make your own information graphics with the programs you are already comfortable using. This graphic is something you could put together in Excel. One of the common questions I hear goes something like this: “I want to use more infographics but HOW do I make them?” I often use the Adobe Suite to make my graphics, but sometimes Excel can be a decent tool for making fairly sophisticated tables. I would not recommend trying to use Word to make graphics. You will become so frustrated with the clunkiness of trying to use a word processor as a graphic design tool that you may be tempted to pick up your computer and throw it out the window. Or, if you are a pacifist, to pick up yourself and leave the office for the rest of the day. But Excel is a more robust, stable program that won’t get finicky if you start manipulating cell colors and border conditions.

What needs work

In general, Excel is probably not the program that’s going to generate elegance. It will allow you to use color and line weight to add layers of visual information, but as you can see here, the results are not necessarily going to be attractive.

In particular, this graphic makes weird color assumptions. The red is bad, the gold is good, and though there is a kind of natural spectrum between red and gold, this graphic doesn’t follow it. I would have used a single color and varied the hue. I have no idea why the middle category is grey. In my mind, grey does not appear on the color spectrum between red and gold. To strengthen this table-as-graphic, I’d go ahead and let every cell (except the empty ones) sit on the color spectrum being used to represent the best and worst. Color can be most meaningful only when it is used consistently. As it stands, there is an inconsistency in the middle categories here with the grey and an unnecessary use of two colors where one would have been enough.

I’m on the fence about the use of apparent depth or 3D-ness. The ‘worst’ buttons stick out like red pimples. On the one hand, the wannabe rebel in me is pleased to see that sort of flagrant display. On the other hand, the depth doesn’t so much add information as it adds visual clutter. Red is enough to make the ‘worst’ seem bad, right? I don’t know. Like I said, I’m on the fence. Maybe the depth element adds value because it helps anchor the eye *somewhere* in this rather extensive table. But it’s used so much that I’m not sure that purchase rings up when all is said and done.

Overall, presenting tables-as-graphics introduces an information overload scenario, one that this particular approach did not surmount. But that doesn’t mean all tables are bad or all uses of color in tables is bad.

I am also deeply skeptical about the Gallup Global Well-Being Index. I’d skip it. Who the heck knows what it means to have a failure to thrive? Very skeptical…

References

Blow, Charles. (2011) “Empire at the End of Decadence” in The New York Times, 19 February 2011. Featuring information graphic “American Shame”.

Trends in returns to college degrees, 1973-2009

What works

Looking at change over time is often best when using simple trend lines. They are easy for the eye to follow – easier than if the same figures were depicted as bar graphs. Given that there are measurable and meaningful differences between the returns for men and women, it is a good idea to show two separate trend lines, as they have done here.

What needs work

The major problem is that returns to college education do not come only from the education received. This trend line is a simple construction that cannot sort out how much income can be attributed to college alone. Sociologists know that a combination of factors – from parents’ educational attainment and parents’ income to things like the student’s aptitude – impact measures of the student’s attainment (like income and wealth). A far more sophisticated model would estimate just how much income one could expect, all other things being equal, for each additional year of schooling. That’s a much tougher model to construct and it wouldn’t be something that could be plotted using trendlines.

In fact, one of the big problems with trend lines is that they are often overly simplistic. On the other hand, they can be excellent representations of the big picture, whatever that might be. There is no simple rule I can think of that would help sort out when a trend line is a great idea and when it is overly simplistic. In this case, change over time is hardly the main story. The real wrinkle when it comes to education is that it can be difficult to determine if students are receiving indoctrination into social networks, ways of acting, and professional networks while they are at college and that these are the advantages that lead to the later bump in income or if they are receiving important knowledge that makes them better, more qualified workers. What’s more, even if we will never be able to divorce the networking from the knowledge gained, we still wouldn’t know how much the background a student starts school with influences their later life choices. Think about this. If someone deeply embedded in a network of people who would usually be a college attender chose not to go to college but continued to hang out with the same people and therefore received much of the same college experience, social and professional networks, how would they fare later in life? Since social scientists cannot randomly assign some students to attend college and others not to, it is very difficult to answer this question. And in this case, a trend line is an oversimplification that misses the major questions about returns to higher education altogether.

Knowing what we know about the various influences on wages later in life and what we see in the trend line, we might assume that women are better able to use educational attainment to escape lower incomes that would have been predicted by, for instance, their parents’ education and/or income. But again, the suggestion that educational attainment has some kind of positive influence on wage premiums is correct, but incomplete. Any assertion about the relationship represented by this graphic is likely to be inaccurate and certain to be incomplete.

References

Blau and Duncan’s Status Attainment Model.

Human Development Index Map

Massachusetts Human Development Index

What works

Mapping the Measure of America is a social science project that deliberately includes information graphics as a communication mechanism. In fact, it is the primary tool for communicating if we assume that more people will visit the (free) website than buy the book. And even the book is quite infographic dependent. I support this turn towards the visual. I also support the idea that they hired a graphic designer to work with them. Often, social scientists do not do well when left to their own under-developed graphic design skill set. Fair enough.

The website presents a unified view of the three images above. I couldn’t get them to fit in the 600 pixel width format, so I presented them one at a time. I encourage you to go to the website because one of the greatest strengths of this approach is the interactivity and layering. I happen to have picked Massachusetts, but each state plus DC has it’s own graphics available. There are other charts and whatnot available, but I think that this set of graphics (which you see all at once) are the strongest.

What needs work

Maps. Maps are too often used. Here’s why I think maps are a problem. Look, folks, political boundaries are meaningful when it comes to making policy or otherwise dealing with state-based funding. And that’s about it. Political boundaries occasionally coincide with geographical boundaries, but not always. Geographical boundaries are meaningful for some things – life opportunities may be based on natural resources or on historical benefits accruing to natural resources. But political boundaries and maps are often not all that useful because they imply that the key divisions are the divisions between states or counties or neighborhoods. Like I said, sometimes this is true because funding tends to be like the paint bucket tool – it flows right up to the boundaries and not beyond, even if the boundaries are arbitrary or oddly shaped. But where the issues are not heavily dependent on funding, thinking in terms of political boundaries makes it harder to see patterns that are organized along other axes. For instance, I wonder what would have happened if some of these categories – education, longevity, income – had been split between urban, suburban, and rural areas. Or urban and ex-urban areas if you prefer that perspective on the world as we know it.

In the end, I think the title is both accurate and disappointing: “Mapping the Measure of America”. Figuring out how to do information graphics well means figuring out which variables are the key variables. In this case, it seems that the graphic options might have determined the display of the information. Maps are easy enough – they appear to offer a comparison between my local and other people’s local. Those kinds of comparisons offer readers an easy way to access the information because everyone is from somewhere and there is a tendency to want to compare self to others. But ask yourself this: to what degree do you feel that state-level information is a reflection of yourself? Do you see yourself in your state?

References

Burd-Sharps, Sarah and Lewis, Kristen. (2010) Mapping the Measure of America with the American Human Development Project. Site design credit goes to Rosten Woo and Zachary Watson.

Admissions rates at top schools drop | Yale Daily News

What Works

The above graph was produced by Yale Daily News. It is clean and does a good job of displaying their admission status compared to their competitors. The reason I thought it was worth mentioning is that a few small aesthetic decisions make the graph pleasing. I like the open circles. I like the fact that the ending values are included as numbers. I would have liked it if they had included starting numerical values, too.

Comparison

For those going through the college admissions process, it can be all-consuming. The New York Times runs a blog called The Choice that focuses solely on this process from the testing to wait lists to moving, transferring and everything in between. Unsurprisingly, then, they ran a table showing similar information about a larger number of schools which they gathered through a mix of old-fashioned reporting – contacting schools and asking them – and Web 2.0 reporting in which schools who had not made the initial deadline could email their data in to be added to the table. Have a look below.

The Choice blog at New York Times 2010 admissions data | J. Steinberg

Ask yourself about the difference between a table and a graph when it comes to conveying information. Edward Tufte is a fan of tables because they can display a great deal more information than a graph. That is true in this case – look at how many more categories of information there are in the table. What do you think? When is it better to present a table full of all the details and when is it better to display a graph like the one above?

References

Lu, Carmen. (5 April 2010) Admissions game getting riskier. Graph. Yale Daily News.*

Sternberg, Jacques. (2 April 2010) Applications to Selective Colleges Rise as Admission Rates Fall. The New York Times “The Choice” blog.

*Note that I wonder if the graphic designer got the data from The Choice blog piece – the publication dates could just be coincidental.

Where does my money go? in the UK - Open Knowledge Foundation, raphic by Iconomical

What works

This visually arresting graphic does a great job of presenting data about national spending in an apolitical but altogether fascinating way. It’s interactive, by the way, but I’m not commenting on the interactive part, just the static graphic. I find that getting the static graphic clear is an important first step towards making a functional interactive graphic. If ever I hear someone say ‘but it’s interactive’ as an excuse for having a weak static graphic, I cringe. See my post about the USDA mypyramid food guide for a case study on the importance of a strong relationship between the static and interactive iterations of graphics as tools.

Each dot represents a different department or governmental program with the size corresponding to the funding level. Smart.

If you link through to the originating site, you’ll be able to follow blog posts that take readers through the development of the graphic. They ask for input and do their best to incorporate it. I like that approach. Good use of technology, OKF.

What needs work

I can’t quite tell why the circles are arranged the way they are or why their hues are the shades they are. Graphics, especially the beautiful ones, are the best when their simple clarity gives way to an elegant complexity. In other words, when I pose the question: “why does the hue vary within given funding types?” I’d like the graphic to lead me to an answer. I’m sure there is a reason for each hue, I just haven’t been able to figure it out.

One tiny, American-centric request: Add ‘UK’ to the page or the graphic somewhere. Maybe change “Total spending” to “Total UK spending”. Or “Where does my money go?” could be “Where do UK taxes go?”. These here interwebs are global. Yes, of course, the £ symbol tends to give it away. Maybe I’m just being too picky.

References

Open Knowledge Foundation. (2009) “Where does my money go?” United Kingdom. Data available