I actually think this infographic is ok

Under the heading, “bad charts,” Mark Duckenfield links to this display by Quoctrung Bui and writes:

So much to go with here, but I [Duckenfield] would just highlight the bars as the most egregious problem as it is implied that the same number of people are in each category. Obviously that is not the case — the top 1% and the 90-99% group, even if the coverage were comprehensive which it isn’t, would have fewer people in them than the decile groups.

But even more to the point, there is no reason to think that the top 10 jobs in each category all yield the same total number of jobs since there much be dozens, if not more broad categories of employment, most of which are off. They’d have been better to have a big “other jobs” block at the end so the bars balanced out. But I suspect the coverage of the top 10 jobs in each category is under 50%,so you’d see a lot of “other”

And this leads to the implication about total number of people making certain incomes might be the same. But all that is the same is their percentage of the employment among the top 10 jobs in the income range.

And I’m not entirely sure that the median salary, which looks to conveniently be $40K is correct. Household income is more like $50K and the median wage earner had a *median* net compensation according to Social Security of around $27500 (although I expect that has deductions and other withholdings from SS wages like health insurance). And the *average* net compensation was about $42500.

My reply: I agree that these graphs have problems but I kinda like them because they do contain a lot of information, if you don’t over-interpret them.

8 thoughts on “I actually think this infographic is ok

  1. Any graph would have “some” problems.

    This is the sort of criticism that makes me skeptical of graph critiques unless the criticizer offers his version of a “better” graph for us to compare against.

    Graph criticism is just too easy a pasttime.

    • Rahul:

      I disagree. As a person who makes graphs, I appreciate comments by others. So what if it’s easy? If it’s helpful, why not? And, yes, of course an elaborate criticism with the display of alternatives is great, but that takes a lot of work. If someone doesn’t have the time or energy to do all this, but still wants to give me some quick criticisms, that’s fine too.

    • The problem is that opinion is easy; science – asking a group of representative readers questions that show how much they got out of the graph – is hard
      part of the hardness is that graphs is not really a discipline, but spread out; the only people I know of who actually study user response are some medical people (there is a lady at UMich who does this for cancer flyers for patients)

      i the only one notice where police are ?
      and I bet that doesnn’t include their pensions, which are for life; police don’t have to worry about running out of money; many have this absurd contract provision where pension is based on last 2 or 3 years salary *including overtime*

      it is sort of like veteran benefits: no politician has the guts to say enough already

      • In my experience people are very resistant to doing any empirical testing of what makes a good graph. There’s all sorts of advice floating around but hardly any empirical data.

  2. The graph does seem to be pretty good for doing at least part of what it intends to do (“what the rich, middle class and poor actually do for work”). Sure, it is like a “cross section” of a three-dimensional object. It would be nice to see “cross-sections” in other directions as well. (For example, parallel box-pots of income by job category.)

    And another “snapshot” would be part-time vs full-time jobs.

    And something (this might be difficult) giving information about people who cobble together a living from more than one job — either part-time jobs simultaneously or limited-term jobs (e.g.,holiday jobs; IRS work; administering or scoring educational tests). My impression is that more and more people are doing this. (Plus a lot of people who can’t make ends meet on Social Security or other retirement benefits.)

    Also: one thing that I noticed that might be a mislabeling: “Primary School Teachers” appear at five deciles, but “Secondary School Teachers” do not appear at all. I am wondering if the “Primary” label should really be “Primary and Secondary”.

  3. I hear myself saying “Don’t stratify on Y” to one of my students.

    I actually think sometimes you can learn stuff by stratifying on the outcome, but it requires careful interpretation. In this case, I think we learn something about a combination of mean and variance (and other distributional moments) of income within each occupation category, but bar graphs of mean and variance could disentangle these effects. I mean, I guess I learned that there is a lot more variance in truck driver salaries than I would have thought…

  4. It’s an interesting representation but I wish they had an “other” block because this is telling us what the percentage distributions within the top 10 jobs for each level are. If it was within all jobs it would be a lot more informative because we don’t know if these represent 5% of all people in those levels or 80%. I might have added a narrow separate bar to show that , and had these represented as “bar of bar.”

    I think the data are really interesting and “stratifying on y” makes a lot of sense in this case, because it highlights how “work” means really different things to high income people than it does to low income people. E.g. it’s one thing to say what’s the big deal about raising the retirement age to 68 if you are in the kinds of jobs at the top, quite a different thing if you are in the kinds of jobs at the bottom.

Leave a Reply

Your email address will not be published. Required fields are marked *