Christopher Mihm, associate director for federal management and workforce issues at the General Accounting Office, told the Senate Governmental Affairs Committee Thursday that Census Bureau dress rehearsals have "underscored the fact that the bureau still faces major obstacles to a cost- effective census."
Mihm said although the bureau was correct in changing its procedures for generating address lists, it "has not used or tested them together."
Mihm said motivating the public to fill out its forms and federal partnerships with local governments are also areas of concern, but did say the rehearsal's staffing was "one of the success stories," with turnover rates roughly a quarter of the 100 percent rate often experienced.
Largely ignoring the constitutional questions of the controversial census sampling technique, Governmental Affairs Committee Chairman Fred Thompson, R-Tenn., said the committee must focus on such basic logistical questions. Thompson said: "Frankly, I don't think what we've heard here today is anything at all [like] what we've heard coming out of the Commerce Department. We need to face up to it and get on with it."
Commerce Secretary William Daley earlier this month reported successful completion of the initial phase of dress rehearsals in California, South Carolina and Wisconsin, although House Republicans said it was too early to tell whether sampling had been successful.
Thompson said he thought the courts would resolve whether sampling meets the Constitution's requirement of an "actual enumeration." Ranking member John Glenn, D-Ohio, said he would prefer not to use sampling, and called the estimated additional $750 million needed to complete the census without using the technique a "real bargain," but questioned whether a better count could be achieved.
"If I thought that was accurate, I'd be the first to vote for it," Glenn said. The projected cost of a census using sampling is $4 billion.
Sampling's constitutionality notwithstanding, Mihm told the panel the GAO agreed with the bureau's simulations that showed the technique could, in many instances, yield a more accurate count.
"Our concern is that the proof is in the implementation," Mihm said, adding, "On average and if they can pull it off ... it's appropriate."