For one simple reason: the GOP was willing to throw piles of money at the Census Bureau in the run-up to the 2000 Census, to give the head count its absolute best chance of being sufficiently accurate that the sampling wouldn’t produce a more accurate estimation.
It worked, sampling moved to the back burner even in the eyes of its proponents, and this time around, considerably less resources have been devoted to preparing for the Census, in the absence of a threat of sampling.
A head count will naturally produce a biased result: unless you count a lot of people twice, the difference between your count and the truth is going to be made up of the people you failed to count, which means you’re going to be on the low side.
But to be low across the country, you’ve also got to be low in some specific places. And the catch-and-release methodology of the sampling adjustment is pretty good at identifying the kinds of places that are more likely to be lower in the count than others.
If you add ‘people’ to those neighborhoods based on statistical sampling, you may of course be adding in numbers that do, or don’t, reflect any actual uncounted people in a particular block or neighborhood. But across the stratum containing that neighborhood, you’re going to come in with a count that has equal likelihood of being high or low. Sure, the smaller a piece of that stratum you’re looking at, the more the adjusted count is likely to deviate from the truth. But it’s equally likely to deviate in either direction, rather than consistently sagging below the truth.
silly thing about this whole thing is that even if the do redraw, or add more districts, uh, people still have to vote dont they? If the repubs are worried about losing seats, change your damn message to reach more people, canvas more, run canidiates that minorities would vote for.
It’s not a slam dunk that the dems get more seats, unless the republicans whine about it and do nothing.
The easiest answer to this is to combine the two. Statistical sampling will produce a result and the actual count will produce a result. The two results should be within a certain percentage of each other.
If this doesn’t happen than the actual count should redone and the statistical sample should be re-evalutated till the two come within correct pecentages of each other.
No one says the final tally has to be within a certain time. In the old days it took many years and years to produce a final tally. This should be how it is now.
If the two don’t agree, than it’s obvious that something is amiss and that the census should be redone till it’s correct. The best way to do this would be to due a full census every ten years and a population census every five years like Canada does.
And yes I know they won’t agree exactly but they should agree within a certain few percentage points, be be agreed upon by both parties.
The logical standard of agreement is that they should both yield the same apportionment of representatives in the House, that being the primary purpose for the Census.
And while I can’t find any objection to this proposal on its merits, it does seem like it would be a lot more expensive than the current system.