Pre-testing Makes Perfect

Anyone who does empirical work knows the importance of a pre-test.  This is the time to figure out that a question is ambiguous, that a survey is too long, or that the survey plan will systematically miss certain people.  Katie Porter can tell her tales about pre-testing the 2001 Consumer Bankruptcy Project in a cold Boston courthouse (and nearly giving up on empirical work before she started).

The Census Bureau is no exception. They have to work the kinks out of questions, survey techniques, and the newest technology just like the rest of us.  They get one shot every ten years, and they need to make it right.  Mistakes would echo through federal funding for local economies, representation in Washington, and just about every other study that involves the American population.

So it was a show-stopper for me this morning to learn that the Census may not have the money for a pre-test.  Congress can't agree on a new budget, so it plans to hold spending "level" at the old dollar amounts.  That may sound fair in most cases, but "level" funding means no new money to cover the cost of pre-testing the 2010 census.

There have been some bitter complaints through the years and even some litigation over Census techniques.  Estimations (counting people whom the Census Bureau did not find) has been a particularly hot-potato issue.  Some Census data are just plain puzzling.  Why, for example, does my Statistical Abstract say there are about a million more  married women in the US than married men?  (And the reader who thinks, "Big Love," is watching way too much television.) 

Anyone who has done empirical work has sweated out funding at one point or another.  But this time we should all be sweating.  The census needs money for its pre-test because we all need a good census count. 

Comments

[MT]

I definitely agree. Adequate data is a key to intelligent decisionmaking and its acquisition should not be a partisan issue.