Data visualization matters: Why you almost never want a pie chart

August 19, 2014

The USDA recently released its annual report on child-rearing expenditures, and this has led to the following chart (from the report) being reproduced in various news and blog articles:

Screen Shot 2014-08-19 at 12.46.05 PM

This use of pie charts is offensive to my senses, but not in the way that comma splices make me twitch.  In this case (unlike with most comma splices), it actually matters.  Case in point:  Not long after I reposted one such article to my Facebook, someone commented that the article’s author should have pointed out that the differences were due to women entering the workforce (note: the USDA report does cite this as a reason for the increase in the cost of education and childcare).  And, because of the misuse of pie charts, I could see how a casual observer might think that is the ONLY notable change in costs in the last 53 years.  

Alas, no.  Why not?  Because the pies are not the same size.  So percentages of the pie are nearly meaningless to compare to one another.

Why do you almost never want a pie chart?  The only information that pie charts give you is proportions of a whole.  So they should only be used if (1) the total size of the pools being compared are the same, or (2) all you care about is proportion and the pool size doesn’t matter.  

Displaying the same data as above as a bar chart yields the following:

Childrearing1960v2013_bar

This chart makes it much easier to see the differences in all budget categories in real (adjusted) dollars between 1960 and 2013 expenditures.  Note that while the housing category was roughly equal in the pie charts, there is actually a notable increase between 1960 and 2013 housing costs.  This is because 30-31% of ~$200K is quite a bit less than 30-31% of ~$248K.  This is what I mean when I say that the pies are different sizes.  It’s like giving one person 1/3 of a medium pizza, giving someone else 1/3 of a large pizza, and then just saying that you gave each person 1/3 of a pizza when asked how much food you supplied.    

But wait, you might say, what if I care about proportions as well as real differences?  Well, there is a chart for that too:

Childrearing1960v2013_stackbar

Now you can see the difference in sizes between each of the categories (as well as the difference in totals), in addition to differences in proportions (i.e., each category relative to its total height).   Alternatively, you can use a bar chart to display raw differences and a pie chart (or stacked bar chart by percentage, my preference to pies) when discussing proportions.

In this case, personally, I don’t really see how proportion is particularly relevant.  How much is spent relative to the budgetary whole doesn’t yield much useful information beyond that given by raw differences (while leaving out the important info that raw differences give).  Moreover, displaying the data ONLY in terms of proportion is downright (though probably unintentionally) misleading.  For example, if one category were to change in real expenditures between time points and all others remain equal, then ALL proportions would change, which would make it seem like things have changed much more drastically than they have.  Let’s say that all that had changed between 1960 and the present were the drastic increase in childcare costs that are presumably due mostly to women working outside the home rather than as unpaid childcare workers.  Our bar chart would tell us this accurately:

Childrearing1960v2013_bar_hyp

Pie charts would tell a different (misleading) tale entirely:

Pie_hyp_1960pie_2013hyp

Similarly, if something DOES change in real dollars but remains a roughly equal share of a whole that otherwise changes quite a bit (as in the case of our actual USDA data with housing costs), displaying in a pie chart obscures that change.

Moral of the story?  Choose your charts wisely.  There are lots of guidelines out there for visual display of quantitative information.  And when in doubt, avoid the pie.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: