Saturday, September 11, 2010

Lies, Damn Lies and Statistics

One of the things I discovered today while poring over data from the BEA, the BLS and the Fed is that there are many, many ways to lie with statistics.

The Fed relies upon plausible deniability.  It merely aggregates data from the BEA and BLS with its own data, and leaves the reader to figure out how credible that data really is.  The adjustments I've found are troubling (seriously, how can you revise outstanding data 30 years after the fact?) but minor.

The BEA relies on revisionism.  It just takes the same underlying data and pushes it up or down at will in order to provide top line growth numbers in line with expectations.  But at least it provides the underlying the data.

The BLS just flat-out lies.  How?  Where the BEA insists on revising ALL historical data when it employs new methodolgy, the BLS does not.  As much as the BEA distorts reality long after the fact, you still have an apples-to-apples comparison (hopefully).  What the BLS does is change the function for calculating U3 unemployment but maintains the fiction that the funciton is unchanged, presenting the data as a single series.  Whenever you graph data, you implicitly assert that the same function is used throughout the graph.  John Williams documents how the BLS has changed its methodology over the years here.  In essence, the BLS makes an apples-to-oranges comparison but pretends that it is only comparing apples.  That's called a lie.

The net result of these three approaches to lying is an overstatement of the health of the economy.  As I start rolling out my findings regarding the data, you'll understand why it matters.