Comments on: Inequality Bar Chart

By: Jon

Jon — Thu, 05 Dec 2013 00:58:39 +0000

Kim,

In response to your question, Tim emailed me the following: “The chart is for college students in the social sciences interested in inequality and social distance. The chart has appeared in many textbooks and Professors have told me they can build a whole lecture around the country differences (top, bottom, 90-10 , Gini) and what they mean. It can also be useful with a brief explanation of one page or less for regular people–meaning the college educated. I have used it at Swarthmore, Simpson, Central Michigan, SUNY Genneseo and elsewhere for undergraduate lectures, and the students all understood it.”

On a personal note, Tim was my advisor in graduate school and I remember creating these charts for him long ago. It’s ironic that this is the first graphic on this site.

Jon

By: Matt Fletcher

Matt Fletcher — Thu, 05 Dec 2013 00:44:20 +0000

A thought just popped into my head – proportionately Mexico is roughly “equally unequal” both ways – ie the median earner gets 3x more than P10 and P90 gets 3x more than the median. But the length of the “median to P90″ part of the bar is much longer (3-1 = 2) than the “P10 to median” part (1-0.33 = 0.67).

It will depend on what exactly Tim’s trying to show, but I wonder whether if the focus is on the length of the bars, it might be better to somehow base this on the reciprocal of the P10 figures?

By: Kim Rees

Kim Rees — Wed, 04 Dec 2013 22:17:13 +0000

Ah, thanks for that explanation. That makes total sense.

So my next question is for Tim. Who is the audience for this? Economists or regular people (or someone else)?

By: Matt Fletcher

Matt Fletcher — Wed, 04 Dec 2013 22:14:21 +0000

Kim – I couldn’t say for sure but I think it’s like this:

Assume you had 100 people in a country, and lined them up in earning order from lowest to highest. Then the median (P50) person would be 50th in line, P10 would be the 10th person in the line and P90 would be the 90th. To get the figures in the table, everything is divided through by the median earnings.

So taking the American distribution, if the median person earned $40,000pa, then P90 would earn 40 x 210% = $84,000pa (and 10% would earn more) and P10 would earn 40 x 39% = $15,600pa (and 10% would earn less)

By: Matt Fletcher

Matt Fletcher — Wed, 04 Dec 2013 22:03:06 +0000

I’m not so keen on joining together the countries as they’re not directly linked to each other. How about something like this (I think this is roughly what Bryan was suggesting?)

I did a version adding horizontal gridlines but I wasn’t sure they were needed.

By: Kim Rees

Kim Rees — Wed, 04 Dec 2013 22:00:40 +0000

I’m still trying to wrap my head around what the P10 and P90 numbers are. “Percent of Median?” What? So the 90th percentile figure (say 210 for the US) is the percent of the median top 10% of incomes? That makes no sense, right? Can someone unravel what those values mean? Then I think I can offer some critique/advice.

By: Ramnath Vaidyanathan

Ramnath Vaidyanathan — Wed, 04 Dec 2013 16:00:35 +0000

I would flip the axes and then do a line plot, one each for P10, Gini x 100 and P90. This is tufte-esque and conveys the key messages very clearly.

By: Tim Brock

Tim Brock — Wed, 04 Dec 2013 14:30:34 +0000

I generally agree with Bryan’s post.

I’m also not a fan of the outer box around the bars. It doesn’t really serve a purpose but encloses the bars away from the rest of the data making the reading of the table from left to right more of a chore than it could be.

By: saleiva

saleiva — Wed, 04 Dec 2013 13:59:13 +0000

I agree with Bryan, the bars doesn’t look like gaps. Also, ordering by gap would remove some noise when visually scanning the composition.

By: Christopher Ingraham

Christopher Ingraham — Wed, 04 Dec 2013 13:44:52 +0000

I second the dumb-bell chart idea. I’d also consider reporting on the figures as concrete dollar values – “$87,000″ is much easier to wrap one’s head around than “179% of the median income.” I might also ditch the years from the country labels in favor of a blanket footnote along the lines of “Date of data collections differs by country. See XXX for details.” Finally, I’d order the bars by lowest to highest gap, or vice-versa.