I have always been a big fan of New York Times especially when it comes to creating cutting edge graphics. A few are mentioned in some of my previous posts. NY Times – Home Foreclosure Graph and
The rise and fall of movies – NY Times.
However once in a while, they err and publish..well…lets say less than world class charts. To my dismay one such chart was published recently:
The charts makes an effort at tracking the change in prices in selected cities over an eight year period in monthly intervals. A causal observer would tend to think that the height of the bars represent the change in prices for that period. However a closer observation reveals that the figures on the vertical axis cannot be for that particular month – they range from 0 to 50% – obviously not a monthly rate of change for an asset class like real estate.
If you look a little more carefully, it may strike you that the % figures may have been measured for that period and shown as compounded for the entire year. Even then….the reader is prone to making mistakes…when the chart slope moves down from 50% all the way to 0%, one may falsely presume that the prices had actually had started to decline. Or having to guess the previous point when the first bar reads say 10%……..10% over what – 100% or 2%? The problem arises when one tries to complicate charts by trying to plot a derivative (rate of change) of a base factor (Shiller Index in this case), while representing the base factor itself would have been sufficient.
A different representation may have been simpler and easier to understand.
Another option, one that works well when comparing two or more plots, is converts all lines to common-size (process of converting all figures to a % of the starting point), like the one below. The obvious disadvantage is that the information about the absolute value of the variable at any given point in times is lost. However it is nonetheless useful – if you had a portfolio of equities (oops I apologize…I meant to say bonds), you may be more interested in how they performed relative to each other and the index than in the absolute size of the portfloio they represent.