I remember Freakonomics. Thing is, it all depends. There is a
   LOT of very useful information that can be drawn from large
   scale studies over long periods of time, given accurate data
   collection of course.

   Freakonomics is basically statistical analysis gone wild with
   correlations galore. It's fascinating and it's useful and _can_
   have predictable capabilities *if* used properly.

   The true test is to look at past predictions to see what held up
   and what didn't.

   Like any predictive method, knowing the limitations of the
   system is critical. One erroneous hypothesis will lead to
   erroneous conclusions, no matter how marvelous the data
   inbetween is. But that's true of any scientific endeavor.

   Also important is that numbers are analogies to real world
   things. They represent real world things but they're not real
   world things.

   At the same time, finding similar flocking behaviors is very
   useful. There's a strong measure of predictability when doing
   cross-discipline comparisons and THAT'S perhaps where something
   like Freakonomics really shines... just as long as it
   continually compares itself to the non-numerical things it is
   representing to check for accuracy. Oftentimes an elegant graph
   continues being shown as if true, when it's not, because the
   story the graph tells fits the story we want to tell.