Lying with Pretty Charts: It’s Still Lying
There are probably as many ways to lie as there are liars. One form of lying that particularly annoys WC is through deceptive charts. Blogger Hentry Blodgett at Business Insider has inadvertently provided a handy set of misleading charts that illustrate WC’s point.
We’ll start with one that is only a little deceptive: Unemployment rates:
Civillian Unemployment Rate, 1940 - Present
The grey areas in the chart are official periods of recession. WC would describe this graph as only slightly deceptive because the “Y” axis – the vertical bit – goes to two, not zero. The compression of the unemployment rates only slightly exaggerates the appearance of the changes in the rate.
By contrast, this chart is a whopper of a lie:
Wages As a Percentage of the Economy
If you examine the “Y” axis of this chart, it in fact ranges across just 11%, from 43% to 54%. No doubt it shows a serious decline in total wages divided by the total economy (however those may be measured), but if viewed on a 0% to 100% chart this would be a gently sloping line, not a graphic grand canyon.
Doubtlessly, the St. Louis Federal Reserve Bank and Mr. Blodgett would argue that by compressing the vertical axis, they better illustrate the problem. WC’s problem is it makes the graph deceptive and misleading.
Don’t get WC wrong. The concentration of wealth, the decline in true earning power and the unemployment rates are all terrible problems. WC has written at length about the amazing power of the Neocon to make the chickens clap for Col. Sanders. But we shouldn’t make our points with deceptive graphics. It weakens our position.
“Make your charts log-log or ln-ln and you always can find a trend out of your data”, or some such tongue-in-cheek advice comes back to me from grad stats&prob.
Appropriate here in that this blog entry appears to me to focus strictly on the initial visual impact, which I aver is specious, as I hope the following example will clarify:
WC: Please graph the change of atmospheric CO2 as a percent of total air over the past 200 or so years, by the criterion you lay out – ie, the y-axis on a 0 – 100% scale. Bingo! There obviously is NO problem!
So……
alaskaranger
October 18, 2011 at 9:50 am
That’s one of the first things we learned in college stats class. Our teacher referred to them as USA today charts. But most ‘news’ outlets do the same thing to impress their political or ideological point of view. I find that nearly all published stats, polls and comparative analysis are horribly flawed. And usually the flaw is in the very basis of data and alot of people don’t even realize it.
One of the most simple examples – ‘Rate your favorites from a list and give them each a score of one to ten’
You add up each choices’ points and divide by the number of votes and the choice with the highest score is, are you ready, almost without exception, simply the least objectionable. Not the most liked. Again, depends on the size of your data set of course, if only three people are voting you have a better likelihood of ending up with the ‘most liked’ but the larger your data set, the further your choice travels toward the least objectionable.
Math is hard. Charts are easy…..
mrderik
October 18, 2011 at 12:46 pm
Sorry to disagree, WC (I seldom do with you). Many charts are not zero-based. In fact, when talking about percent unemployment, you are already not using the raw numbers (number of unemployed). Dividing by the size of the work force makes it more understandable. And we understand numbers like 8 to 10, better than 14,000,000.
Showing the daily temperature on a chart looks fine, if the chart scale is from the low to the high (today in Pasco, say, use a graph from the 40′s to the 70′s, and see the squiggles). But as every scientist knows, today’s temperature only ranged from about 280 to 294 K. That makes a poor chart, but it wouldn’t be “Lying with Pretty Pictures”. I bet you’ve never looked at an Alaskan climate chart that was absolute zero based. People like 0 degrees F and 0 degrees C because they are within our “normal” experience.
One last example – almost all financial charts use the range over the time of the chart. A 1 day chart of Apple’s stock will only show about the day’s range, while a 20 year chart will use a much larger range. The 1-day chart would be hard to read if it were zero-based.
PascoBill
October 18, 2011 at 9:05 pm