Presenting Data: Tabular and graphic display of social indicators
Gary Klass
Illinois State University

Note: The website will be discontinued shortly, to be replaced by the Just Plain Data Analysis site

Home Up Good Tables Good Charts Divide Analyzing Budgets Election Data Education Poverty References Course page Chart of the week your comments

Bad Charting:

The Charts used here to demonstrate bad charting practices are taken from the "Illinois Higher Education Annual Report" prepared by the Illinois Board of Higher Education, July, 2002. 


Problems with this chart:

Displaying the total enrollments on the same chart as the Community, Public and Private doubles the scale, making it very difficult to see significant differences in enrollment growth rates between the private and public institutions.  Nevertheless, because this is the first chart in the section of the report profiling higher education in Illinois, total enrollment figures are a key indicator that should be reported in the first table.  One solution would be to lengthen the vertical axis, as was done in an earlier chart contained in the 2001 IBHE Data Book:

Note how the increasing private sector enrollments are much more obvious in this chart.  The second chart also avoids two minor shortcomings of the first: the unnecessary gray background and the use of curved lines between the data points.  This chart is not as pretty as the first, but it shows a more accurate depiction of the data.  Among the problems that remain are gridlines that are too thick and the "over 30 years" text in the legend.

Whereas the first chart reports total fall enrollments the next chart, from the 2002 Annual Report, reports 12-month headcounts (a larger number, representing the number of individual students enrolled at anytime during the year).  Although the data are very similar to those shown in the first chart, the authors of the report chose to use a different chart type and make some very bad chart formatting decisions.  The result is a chart that deserves at least honorable mention as one of the worst charts ever constructed:

The first obvious problem here is the use of the unnecessary 3-D effect.   It takes a couple minutes to figure out whether the Y-axis scale references the front or the back of the bars: is the first "Community Colleges" bar greater than or less than 700,000?  Note also how much more ink is used in the bar for the private institutions data than is used for the public universities data.  Placing the Community College bar in the middle of the three sets of bars seems an unfortunate choice.

But wait!  There's a less obvious but far more serious problem with this chart.  As you examine the trends in these data, the first impression is that the Community College bars are increasing and that the public university enrollments are catching up with private institution enrollments.  But this is just the opposite of what we saw in the first chart.  Look again  -- at the tiny numbers representing the years at the bottom of the chart.  The chart is backwards! Always depict time from left to right.

Interestingly, the IBHE did a better job -- but still not good -- of presenting these data in their previous years' Annual Report:

This chart shows a better placement of the Community College bar and the years are in the correct order.  It still suffers from the unnecessary 3-D effect, but at least gridlines allow a better visual estimation of the data.  Note also the awkwardness of the white border around the black bar and how the border does not show up in the legend.  Note that in contrast with the 12-month headcount chart, above, in order to estimate data points for the bars, you compare the back of the 3-D bar with the gridlines.

The distorting effects of 3-D graphics used throughout the Annual Report are best illustrated in the pie chart on the left.  See how the angle and placement of the pies serve to exaggerate the size of the public university slice and how the white border around the community college slice reduces its size.

Generally, pies do not provide for an accurate visualization of data, as can be seen in a comparison with a bar chart representation of the same data (on the left).

The next chart illustrates the problems that frequently occur with stacked bar charts when the ordering of the elements on the stacked bar is arbitrary.  Again, the 3-D effects only compound the difficulty of visually estimating the relative size of the components of the bars.

The Importance of sorting data.

Sorting data by the most significant variable greatly aids in the interpretation of graphical representations of data.  Sorting data alphabetically, as is done throughout the IBHE report, generally works to hide significant facts about the data.

Problems with this chart:

  • The legend at the bottom of the chart shortens the plot area.
  • Unnecessary 3-D effect.
  • Unnecessary shading of plot area.
  • The white borders on the undergraduate bars make those bars appear shorter; the white borders do not show up on the legend.
  • The universities are sorted in alphabetical order.

Correcting these problems results in a chart, below, that says more about the data. 

While sorting data alphabetically is a common graphical mistake, it is  even worse to have the data sorted randomly, as is the case in the next chart, depicting the number of degrees awarded by major.  Here the data are sorted not alphabetically, not by the 1980-81 data, nor (as would be best) by the 2000-01 data. 

What explains this unusual sequencing of data?  The answer, it seems, is "consistency", as in Ralph Waldo Emerson's observation, "A foolish consistency is the hobgoblin of little minds".   In this case, the IBHE authors are maintaining a consistency with their previous years' report, where the following chart appeared:

Here we see the top majors sorted at least in some order, although the 3-D effect makes it difficult to figure out whether Psychology's bar is larger than Bio/Life sciences, we will assume that the Psychology bar is the larger of the two (note again, with this 3-D effect, you have to use the back of the bar).

So the original chart displays the number of degrees for the top ten majors in 2000-01 and 1980-01 sorted by the number of degrees in 1978-79.  This, however, raises an interesting question: In what years were these the top ten majors?