Ethics: October 2010 Archives
A couple of days ago, I posted about this silly graph, which shows the wage gap between men and women:
The dotted gray line on the graph at first seems to show that "Women's Wages as a Percentage of Men's Wages" are dropping, but that turns out to be because the dotted gray line is plotted against the right-hand scale, which is printed upside down, with the larger numbers at the bottom. I called this a sin against graph design.
My co-blogger Ken suggested in a comment that this might not be such a terrible sin:
I often run into a graph like that where the right-hand scale is reversed when reading scientific papers. It's often used to emphasize a reduction such as this one. The graph clearly shows an overall reduction in the gap between wages and demonstrates the variations by time.
It's actually a good graph, but fails to make the point James Park was trying to make. The question is if James Park created the graph with the intention of deceiving, or just copied the graph from an analysis done that Park failed to understand.
A little research indicates it was the second option. The graph appears to be Figure 8 from the report "Women and the Economy 2010: 25 Years of Progress But Challenges Remain" dated August 2010, which appears on the "Women and the Economy 2010 Series" website. The report is credited to "the Majority Staff of the Joint Economic Committee" of the United States Congress.
So, it's your tax dollars at work. I imagine James Park probably just took the graph from that report without worrying too much about the details. After all, it's from Congress, right?
In any case, it's still a bad way to show data. I realize that data sometimes has to be massaged to show results--such as plotting data against a reciprocol or logarithmic scale--but inverting the scale like this just seems wrong, especially since you could just as easily plotted the size of the gap itself, which really would slant downward on a normally-oriented scale.
I suppose one possibility is that for some historical reason wage differential graphs are always presented in this format. That sort of thing sometimes happens. For example, the traditional way to draw supply and demand curves in economics is to put price on the vertical scale and demand on the horizontal scale, which swaps the dependent and independent variables from the usual scientific practice. However, since it was an economist's blog that first brought his to my attention, this seems unlikely.
If you know anything about graphic presentation of data, especially if you've read Edward Tufte's Visual Display of Quantitative Information
A quick glance shows that men's and women's wages are graphed against the left-hand scale. As you can see, both show a slight rising trend in constant dollars. The gray dashed line shows women's wages as a percentage of men's wages, and it's declining, which seems to show a loss of parity for women. This is odd, because the colored wage lines above are visibly getting closer together. Also, we may not be living in a paradise of equality, but does anyone who's been paying attention for the last 25 years really think things have gotten worse for women?
The explanation for this mystery becomes apparent when you realize that the gray line showing women's wages is plotted against the right-hand scale, which is printed upside down. Women's wages are in fact rising slightly faster than men's wages, so if the scale was printed correctly, the graph would be slanting up. Unless you resort to outright fabrication, data presentation doesn't get more deceptive than this.
[Update: More here, including James Park's apparent source for that misleading graph.]
(Note: Such a broad-based comparison between men and women is nearly useless anyway, even if printed correctly, unless you control for things like age, education, the type of job, and years of experience.)
(Hat tip: Steve Landsburg)
.
