TODAY'S HEADLINES
HOME PAGE
SECTIONS
• NEWS
• SPORTS
• BUSINESS
• LIFESTYLES
• NEON
• Dining
• Showguide
• Nightlife
• Transportation
• Movies
• TV Listings
• OPINION
• Columnists
• IN-DEPTH
• CLASSIFIEDS
• WEATHER
Recent Editions
• Monday
• Tuesday
• Wednesday
• Thursday
• Friday
• Saturday
• Sunday
• Archive
Fun Stuff
• Crosswords
• KIDS AREA
• Gallery
• Buy Tickets
• LV Links
• Site Index
• Newspaper Subscription
• News Tips
• Contact Us
• Corrections
• Rate us
|
Tuesday, July 28, 1998
Squabble over 'sampling'
Census should be a count, not a projection.
Congressional Republicans now threaten to withhold or limit $952 million in money earmarked to prepare for the Year 2000 census, in a dispute over administration plans to use computer "sampling" to supplement or (in some cases) replace more traditional methods of head-counting.
"Sampling" is a process by which statisticians gather detailed data about a limited group of people, and then extrapolate that data onto a state or nation.
Both sides of the debate seem to believe such computer extrapolations will increase the count of racial minorities and the economically deprived, particularly in urban areas.
The purely partisan concern is that beefing up inner-city population counts via "sampling" may divert more federal funds -- and perhaps a few more congressional districts -- to urban areas, which tend to vote Democratic.
But that's the wrong way to look at the issue. If more blacks or Hispanics -- or Eastern Europeans, for that matter, or homeless hobos -- can really be found, they should indeed be counted and represented.
The problem is that sampling, by its very nature, takes any error in the initial sample -- or in the process by which the sample populations are selected -- and multiplies it.
Although the result will tend to look far more "scientific" than a number picked out of a hat, programmers have long referred to this effect as GIGO -- "Garbage In, Garbage Out."
Let us suppose that errors creep into a sample count of the population of Corpus Christi. Large numbers of undocumen- ted immigrants appear to be present in the metropolitan area, though many run away or hide from the census-takers. In the end, the census-takers' "best guess" is that the Texas city of 300,000 has an Hispanic population of 100,000 (somewhat higher than the Texas state average of 26 percent), and that there are 20,000 undocumented residents -- a guess which happens to be off by 12 percent.
The computers are then set to whirring, and dutifully inform us -- down to a specific-sounding decimal point -- that Wisconsin, with a population of 4.9 million, must have 1.63 million Hispanic inhabitants, of whom 326,667 are undocumented seasonal agricultural workers.
The example is exaggerated -- hopefully the Census Bureau knows there aren't many avocado groves in Eau Claire. But the fact is, the Constitution calls for an "actual enumeration" -- a counting. If we're going to rely on computer projections, why not really save some dough? Just run out existing trends for another decade, and issue the 2010 census at the same time?
Americans are counted and tracked today as never before. Working from birth and death certificates, school and vaccination records, voter registrations and motor vehicle data, Washington should be able to get a line on 96 percent of us, pretty quickly.
For the 4 percent where questions remain, let the census takers go forth and count, in person.
Republicans consider sampling an untested and unreliable science. Americans elected GOP majorities to control both houses of Congress -- and with them, census funding. Therefore, sampling should not be used.
Give us your FEEDBACK on this or any story.
Fill out our Online Readers' Poll
|
|
