January 28, 2008

How to mislead with charts

At the end of last year, I wrote a review of several Friedman Foundation reports that claimed voucher programs would increase graduation. You can read my full remarks from the link above, but let me focus on a graph that appeared in the South Carolina report:

Original-Gottlob-chart.JPG

For the moment, let's assume that the point estimates are correct (They'd be biased somewhat by the fact that the Current Population Survey does not survey military personnel or those who are in jails, prisons, and other institutional settings.) Look at the numbers to the right of each bar, and then look at the bar lengths.  You'll see that though the numbers indicate that there were fewer dropouts estimated than college graduates or those with some college, the bar for dropouts is longer.  And also look at the scale, which has equal intervals at the bottom for 0-100K, 100K-300K, 300K-600K, and 600K-900K. I have no idea how the author constructed the graph (despite the how in the title of this entry), but below is what a less misleading graph would look like (if with my spectacular-NOT skills in producing a clean JPEG from an Excel chart):

Revised-Gottlob-chart.JPG

The corrected chart shows that using the author's figures, high-school graduates far outnumber those in every other category. That fact doesn't mean that we should sit on our hands about the students who don't graduate from high school, but it does mean we should be skeptical of the rest of the report.

My favorite sites for discussing such things:

Listen to this article
Posted in Education policy on January 28, 2008 10:45 AM |