Economists often have something interesting to say. However, they often do a terrible job of making their points in a way that those outside of economics can understand.
For some time, I’ve developed an admiration for the approach preached by
I won’t go into Tufte’s approach here, but I thought I’d show how using his approach can radically improve the presentation of economic data. In this example, I’m trying to show Oregon’s unemployment record relative to the rest of the U.S.
This first figure, below, shows unemployment for Oregon and the U.S. as a whole. The chart was made in Excel by highlighting the three relevant columns (month, Oregon unemployment, and U.S. unemployment) and clicking the line chart button.
The result is a standard ugly Excel chart. Even worse, it does not tell an especially interesting story about Oregon’s unemployment. Oregon seems to have slightly higher unemployment than the U.S. as whole, but also seems to track U.S. unemployment fairly closely. Big whoop.
The problem with the figure above is that it does not say anything about unemployment in the other states individually.
The next set of figures use a much bigger dataset. It reflects the unemployment rate for each state in each month from January 1976 through October 2012. That’s more than 22,500 observations.
The figure below is better, but not much better. Oregon’s unemployment is the heavy black line. The gray background represents the range between the state with the highest unemployment and the state with the lowest unemployment.
Notice that the graph removes much of what Tufte calls “chart junk”—a frame around the graph, ugly and unnecessary gridlines, a legend, and the month of “Jan” in the x-axis.
While the graph above is much cleaner and better looking, it is fatally flawed. With a few exceptions, Oregon looks to be in the middle of the range of unemployment rates across the states.
I know this is not true.
In fact in 213 out of 442 months (48 percent of the time), Oregon has been in the top 10 for high unemployment among the states. The figure above does not adequately show this fact.
The figure below is a major improvement. It shows the percent unemployment for each state represented by tiny gray dots, with Oregon represented by the red dots. (I would have preferred a red line for Oregon, but Excel layers dots over lines. Thus a red line would be “under” gray dots and looks amateurish.)
With the figure below, it is easier to see the point that I am trying to make: Oregon’s unemployment tends to be among the worst in the country for many points in time. This can be seen pretty easily in the years 2001 through 2005. But the point is not as well made in 1981 and 1981 when Oregon’s unemployment was the fourth or fifth highest in the U.S.
The figure below is another improvement. The black line is Oregon’s unemployment (the black line is darker than the gray dots, so Excel’s layering problem goes away.) The red dots represent months in which Oregon’s unemployment is in the top 10.
While the graph is much better and makes the point I was looking to make, something about it bothers me and I feel it is cluttered.
The fact that Excel does some weird thing that makes it look like there are white gridlines really bugs me.
Here we go. One last improvement.
In the figure below, I went back to the gray background representing the range between the state with the highest unemployment and the state with the lowest unemployment. The black line is Oregon’s unemployment and the red dots represent months in which Oregon’s unemployment is in the top 10.
The final figure seems to best make the point I was trying to make. We see Oregon’s unemployment, we see the range of unemployment among the states, and we see how often (and when) Oregon’s unemployment was among the worst in the U.S.