A1-HannesHesse

From CS294-10 Visualization Fa07

Assignment 1a: Good and Bad Visualizations

[edit] Bad example of visualization

From: Sinn, Hans-Werner, Can Germany Be Saved?, MIT Press, June 2007

Explanation and discussion

This figure from a recent economics book shows the development of Germany's unemployment over the last four decades. Several factors make this visualization difficult to read.

Around 1990, when the reunification occured, the line branches out into two lines: One (red) line representing Germany's overall unemployment, including East Germany, and one (blue) line representing only West Germany's unemployment. The choice of color emphasizes the figures for West Germany (because they can be read as a continuous line). This is mainly to facilitate an undistorted analysis of the cyclical pattern of unemployment.

The choice of colors and captions are at first confusing: The black caption "West Germany" clearly corresponds to the blue line, while the red, bolded caption "Germany" relatively clearly corresponds to the red line. Because of its similarity to the "West Germany" caption, the caption "East Germany" on the right-hand side of the graph first appears to belong to a line, too. Upon closer inspection, however, the reader finds that it actually labels the interval between the blue and the red line.

So here we have a graph where two (expectedly) comparable figures, West and East Germany's unemployment, are depicted in different ways: One is represented by an explicit line, while the other is not a line by itself, but the gap between two other lines.

The utility of the red line and the gap representing East German unemployment is questionable: The development of East German unemployment is difficult to read from this graph because it depends on the West Germany portion of the red line, but appears to be constant after 1997. Because it is stacked on top of the West German figure, it takes estimation to compare the two figures (if this is at all intended). Even if plotted individually, rather than in a cumulative manner, comparing these absolute figures has limited value because of the vast difference in population between the East and the West. Instead, a greater informational value could be achieved by using relative numbers.

If the main purpose of this figure is to illustrate the cyclical development of unemployment along a linearly increasing baseline trend, the author should have limited the graph to only West Germany to avoid clutter and confusion.

Deconstruction

The chart depicts absolute unemployment numbers in a straightforward manner: The x-axis encodes the years from 1971 to circa 2010 (quantitative). The y-axis encodes the unemployment figure for this year in millions (quantiative, zero-fixed). The line color encodes whether this figure applies to West Germany (blue) or to Germany as a whole (red).

As a consequence of the overlay of the Germany and West Germany curves, the space between them implicitly encodes the unmployment figure for East Germany (quantitative, zero-fixed).

Redesign 1: Leave out East Germany

The first redesign simply leaves out the red line with overall Germany's unemployment, since the text is not concerned with this figure, but rather with the cyclical pattern of unemployment over four decades.

This simple cleanup job makes the chart much less confusing.

Redesign 2: Independent lines for East and West Germany

The second redesign replaces the red overall Germany line with a line for East Germany's unemployment by transforming the interval between the the old red and the blue line into an absolute line.

This figure yields comparable figures for East and West Germany and shows that the cyclical pattern only occurs in West Germany. It is unclear from the data whether the rapid incline between 1990 and 1991 is due to lacking figures from previous years or an actual dramatic increased unemployment in East Germany as a consequence of system change.

Redesign 3: Relative numbers

The third redesign replaces the absolute numbers from the previous charts with relative unemployment numbers (percentage of all citizens in the labor market unemployed). This figure is more robust towards changes in population size and gracefully incorporates Germany's overall unemployment numbers after 1990 without distorting the graph.

[edit] Good example of visualization

From: The New York Times, August 2007

This daily index chart of Standard & Poor's ratings for various companies from the New York Times manages to provide quick access to a variety of quantitative and temporal data.

The position of a company in the X-Y space represents its price return over the last week (X-axis) and over the last year (Y-axis). The choice of these two timeframes may be somewhat arbitrary, but it divides the space up into four meaningful sectors, explained in the caption below:

Leading companies (upper right) have performed strongly, both in a short-term and long-term view.
Improving companies (bottom right) have generally performed poorly, but have managed to yield a positive return over the past week.
Lagging companies (bottom left) have performed poorly, both in a short-term and long-term view.
Slipping companies (upper left) have generally performed strongly, but their price return has been negative over the past week.

Combining these two timeframes into one picture gives an indication that for a meaningful analysis of a publicly traded stock, it is necessary to look at development of key figures over time, rather than to just look at a snapshot.

All of these figures appear to be weighted against a mean price return of the S&P index, and this baseline is easily recognizable by the black crosshair. Because the crosshair is not centered, it is easy to evaluate the general state of the market on a given day (comparing several of these charts over time will show that the center of the crosshair moves around quite a bit).

In addition, the chart incorporates some absolute quantitative data about market capitalization, represented by the size of a company's circle. Unfortunately, the chart does not make clear whether the radius or the area of a circle represents this figure.

Deconstruction

The positions of the circles in the x-y space encode the relative performance of companies compared to the "S&P 500 return": The position on the x-axis indicates the one-week price return while the position on the y-axis indicates the one-year price return. These figures are quantitative.

In a third dimension, the size of the circles encodes the market capitalization, another quantitative measure.