| |
|
Monday, April 2, 2001
|
|
|
The (Ongoing) Partisan Pursuit of Adjustment
|
|
J.Kenneth Blackwell
Recently, a committee of experts at the Census Bureau demonstrated admirable restraint by recommending against adjusting the census based on a statistical sample. Citing major discrepancies between adjusted numbers and population benchmarks, the committee was “unable to conclude…that the adjusted Census 2000 data are more accurate.” This mild wording is a shocking admission from people who had championed adjustment in the face of opposition from Republicans, outside experts and the U.S. Supreme Court. Some of the committee members had pursued adjustment – the statisticians’ Holy Grail – for most of their professional lives.
Unfortunately, the pursuit was not marked by the prudence displayed in the final decision. Democrats in Congress and in the Clinton White House had their own Grail – a population increase in urban areas to favor Democrats in the next round of redistricting. Their partisan agenda coincided with the scientists’ professional one, and an alliance was born. Democrats provided racially-charged rhetoric to demonize adjustment critics. The Bureau supplied a steady stream of “unbiased” analysis that added a professional veneer to the partisan support for sampling.
This resulted in some less-than-objective Bureau pronouncements. In 2000, the Bureau Director began referring to sampling estimates, previously known as “adjusted” numbers, as “corrected” numbers. The change was noted in Congressional hearings and a meeting of the National Academy of Sciences, where independent observers faulted the Bureau for adopting a biased term. Bureau officials politely reverted to “adjusted” in those venues, and consistently used “corrected” with national media. This was propaganda – who could object to “corrected” numbers? – at the expense of scientific objectivity. The result? The committee’s recommendation confounds logic by noting the “corrected” numbers are less accurate than the “uncorrected” numbers.
Or consider the Bureau’s position on subtractions. Any adjustment would add some people and subtract others. Critics anticipated attacking this Achilles’ heel, which might have violated Constitutional protections, in court. Statistically, it is a non-issue – subtracting some people who turned in a census form is bad policy, but acceptable science. So it was doubly strange when, despite claims to focus exclusively on technical issues, the Bureau offered a pre-emptive, tortured, legal defense. The “downward adjustment is accomplished by creating negative weights that, when added to Census 2000 tabulations, reduce the count…No records would have been removed from the Census 2000 files. However, [adjustment] may subtract a person’s individual characteristics from the Census 2000 tabulations.”
Well. This is spin, not science. The simple fact is that adjustment would decrease the population in some areas. As the leading statistical agency in the country, the Census Bureau is expected to report the simple facts, and leave complicated spin to others. Once they began to shade the analysis, Bureau officials compromised their objectivity. If they would re-define “subtract” to make adjustment more palatable, how would they describe other serious problems?
The answer is in the Bureau’s recommendation. The adjusted numbers exceed reliable population benchmarks by millions, which casts serious doubt on adjustment’s accuracy at any level. The report also notes that adjusted numbers are not accurate in areas with less than 100,000 people – which is most of the country. Yet, the Bureau holds out hope that “further research may establish that adjustment based on A.C.E. would result in improved accuracy.” Where, or how? That depends on the definition of “accuracy.” The quest for the grail continues.
As planning begins for 2010, the Census Bureau should put science ahead of spin. In the last decade, the Census Bureau spent over $300 million and countless hours developing an adjustment that proved illegal, impractical, and inaccurate. Those resources could be better spent building address lists, expanding local partnerships, and perfecting management systems – three proven methods of improving the real count of our country, particularly in minority communities. That would be a quest worthy of true professionals.
# # #
J. Kenneth Blackwell is Co-Chairman of the U.S. Census Monitoring Board.
|
|
