10/02: More on Taking the Census
Category: Politics
Posted by: an okie gardener
Last week I commented with suspicion on the move by Obama to bring the decadal census under the control of the White House. Here. Imagine if it were a Republican White House and the Census was reporting to Karl Rove. The MSM and fellow-travelers would be having a fit.
John Fund in a Wall Street Journal essay offers his thoughts.
Here are excerpts:
"There's only one reason to have that high level of White House involvement," a career professional at the Census Bureau tells me. "And it's called politics, not science."
. . .
Mr. Chapman worries about a revival of the effort led by minority groups after the 2000 Census to adjust the totals for states and cities using statistical sampling and computer models. In 1999, the Supreme Court ruled 5-4 in Department of Commerce v. U.S. House that sampling could not be used to reapportion congressional seats. But it left open the possibility that sampling could be used to redraw political boundaries within the states.
Such a move would prove controversial. "Sampling potentially has the kind of margin of error an opinion poll has and the same subjectivity a voter-intent standard in a recount has," says Mr. Chapman.
Starting in 2000, the Census Bureau conducted three years of studies with the help of many outside statistical experts. According to then Census director Louis Kincannon, the Bureau concluded that "adjustment based on sampling didn't produce improved figures" and could damage Census credibility.
The reason? In theory, statisticians can identify general numbers of people missed in a head count. But it cannot then place those abstract "missing people" into specific neighborhoods, let alone blocks. And anyone could go door to door and find out such people don't exist. There can be other anomalies. "The adjusted numbers told us the head count had overcounted the number of Indians on reservations," Mr. Kincannon told me. "That made no sense."
The problem of counting minorities and the homeless has long been known. Census Bureau statisticians believe that a vigorous hard count, supplemented by adding in the names of actual people missed by head counters but still found in public records, is likely to lead to a far more defensible count than sampling-based adjustment.
John Fund in a Wall Street Journal essay offers his thoughts.
Here are excerpts:
"There's only one reason to have that high level of White House involvement," a career professional at the Census Bureau tells me. "And it's called politics, not science."
. . .
Mr. Chapman worries about a revival of the effort led by minority groups after the 2000 Census to adjust the totals for states and cities using statistical sampling and computer models. In 1999, the Supreme Court ruled 5-4 in Department of Commerce v. U.S. House that sampling could not be used to reapportion congressional seats. But it left open the possibility that sampling could be used to redraw political boundaries within the states.
Such a move would prove controversial. "Sampling potentially has the kind of margin of error an opinion poll has and the same subjectivity a voter-intent standard in a recount has," says Mr. Chapman.
Starting in 2000, the Census Bureau conducted three years of studies with the help of many outside statistical experts. According to then Census director Louis Kincannon, the Bureau concluded that "adjustment based on sampling didn't produce improved figures" and could damage Census credibility.
The reason? In theory, statisticians can identify general numbers of people missed in a head count. But it cannot then place those abstract "missing people" into specific neighborhoods, let alone blocks. And anyone could go door to door and find out such people don't exist. There can be other anomalies. "The adjusted numbers told us the head count had overcounted the number of Indians on reservations," Mr. Kincannon told me. "That made no sense."
The problem of counting minorities and the homeless has long been known. Census Bureau statisticians believe that a vigorous hard count, supplemented by adding in the names of actual people missed by head counters but still found in public records, is likely to lead to a far more defensible count than sampling-based adjustment.
