Wednesday, January 14, 2009

How to lie with statistics




The first chart, with the shaded background, is the bogus chart from the Minneapolis Federal Reserve Bank discussed below.
The second chart is the one Spencer at Angry Bear made, showing actual recessions.


http://angrybear.blogspot.com/2009/01/honest-resarch.html

This chart from the Minneapolis Federal Reserve Bank is starting to appear on various web sites. Among others, Alex Tabarrok of Marginal Revolution published it.

It is an interesting looking chart, very similar to many I do.

But it struck me as odd, there was something wrong with it.

I can not remember a single recession where employment did not fall as this chart shows.

So I tried to reproduce the chart and came up with something that looks like this.[obvious differences from the first chart.

Note that it shows the 2001 recession as the mildest recession and it did experience falling employment. This line as well my line for the harshest recession in 1957 are very different than the chart published by the Minneapolis Federal Reserve and referenced by several libertarian/conservative blogs. Their harshest is about a full percentage point deeper than mine.

But the Minneapolis Fed did publish their original work and I was able to determine that there chart was not of an actual recession. Rather their lines representing the mildest and harshest recessions are completely artificial creations that have little or no relation to any actual historic event.

Rather than show the 2001 recession as the mildest recession, they went through all of the first months of the 10 post WW II recession and found the smallest observation and made that the first observation of their mildest recession line. Next they went through the second month of the ten recession and found the smallest observation in the second month of recessions and made that the second month of their mildest recession. They repeated this process for 18 times, each time making the smallest observation of that particular recession month for their line of the mildest recession. So their line might take observation one from 1957, observation two from 1981, observation three from 1973, etc., etc. Their so called harshest recession line was created using the same methodology.

Here is exactly how they described the lines in their publication:

This page places the current economic downturn into historical (post-WWII) perspective. It compares output and employment changes during the present recession with the same data for the 10 previous recessions that have occurred since 1946.

This page provides a current assessment of “how bad" the recession is relative to past recessions. It will be updated as new data are released. This page does not provide forecasts, and the information should not be interpreted as such.

The following charts provide information about both the length and depth of recessions.



Minneapolis Fed

This strikes me as a major case of intellectual dishonesty. At no point is the reader shown that their mildest and harshest recessions are completely artificial creations that have no relation to any actual recession.

The other point that strikes me is how libertarian/conservatives like Alex Tabarrok uncritically accept such biased research from fellow right wing sources without any question. It strikes me that a tenured, PhD, economist should know that there was no post WW II US recession where employment did not fall and should have recognized that there was something wrong with this chart. But more than likely, this chart will now be passed around and used by libertarian/conservatives to demonstrate that there was a post WWII recession where employment did not fall and that will become one of their standard talking points.

Am I being too harsh on the Minneapolis Fed and people who uncritically accept and pass on such biased research?

P.S. This is not the first time I have caught the Minneapolis Fed or Alex Tabarrok misrepresenting data.

================================================================================
Since I posted this article, Alex Tabarrok responded with the following comment on Angry Bear:

Alex Tabarrok says:
Today, 9:51:21 AM [1-15-09]
“I have done some further investigation of the Fed chart and contacted the Fed for clarification.

I was wrong. Spencer was right.

The graphs in the Fed's picture are Frankenstein recessions, recessions cobbled together by taking bits of pieces of each past recession and assembling them to create a mild, median, and harsh recession - none of which ever occurred. I do not think this is a good way of looking at the data - in fact I am shocked the Fed would present data in this way - on this Spencer and I agree.

I certainly was not dishonest about the data but if it had not been the Fed I probably would have taken a closer look.

I have posted a correction and a further post, in which I present all the data, here:

http://www.marginalrevolution.com/marginalrevolution/2009/01/comparing-recessions-ii.html

Alex Tabarrok
================================================================================

I agree with the commenters at Angry Bear that the only reason someone would create such a graph, and label it the way it the way they did, and put in a clarifying explanation where it would be overlooked by many or most, was to deceive people. (Note: Tabarrok did not create this chart, he copied it from the Minneapolis Fed).

No comments:

Post a Comment