2010-09-30

Income Distribution

Matt Yglesias blogged a graph that has been making the rounds across the economics blogosphere today.

I note, however, that these graphs can appear wildly different, depending on the rules one applies to the data in question.

Consider, for example, what would appear in the graph if I a person migrated from one income percentile to another during the nearly 40-year study period (not implausible). Such growth wouldn't show up in the graph. It is merely a comparison of income growth between the same percentile in two different periods. It is *not* (at least, not as far as I understand it) a comparison of the income growth experienced by the same individuals in both periods.

There are four different time periods expressed in that graph:
  1. The 1979-1997 Study Period.
  2. The 1998-2008 Study Period.
  3. The *INDEX PERIOD* for the 79-97 Study Period. This is the time period during which I was determined to be in any particular income percentile. Was this index period Dec. 31st 1979? Dec 31st 1997? The midpoint of those two?
  4. The *INDEX PERIOD* for the 1998-2008 Study Period. The same observations/questions apply here.
As you can (hopefully) see, the graph will appear quite different, depending on how the Index Periods are determined. If my income as of 1979 determines what percentile I am in for the rest of the study, then the graph will look much different than if my percentile is determined both in 1979 and 2008 - or if it is determined by the sum of my income for each individual study period - or if it is determined by the sum of my income for all years from '79 to '08 - or if it is determined only in '08 and applied retrospectively - etc. etc. etc...

These sorts of issues are at the heart of any graph. With income comparisons like these, you can make the story look very different depending on what your aggregation rules are, as applied to the data.

HT: Too many to note...

Addition Note for Today:

I am adding a "Scientific Method" tag. As I continue to explore the economic blogosphere, I am encountering questions that pertain not merely to economics, but to the method by which rules for analysis were established. Today's blog post is a great example. I can think of a few blog posts from the past that dealt with similar issues. It is a pet interest of mine, so I look forward to the new tag.

In general, I continue to refine the labels I have used. I see that I have separate tags for "joy" and "happiness," for example. Clearly, this was an unintentional redundancy. I will endeavor to clean that up a little bit, going forward.