RecountsThey always change
By Ed henry
RecountsThey always change
This Florida recount situation is bringing back all sorts of memories for me. Having spent so much of my life in the market research business and having played the numbers game for quite some time before pioneering qualitative exploratory research, I feel great sympathy for those now hand counting ballots and running IBM cards through sorters time after time.
The vote-counting procedures or technologies used by much too much of the nation are from my era. Theyre from the 50s when I was a numbers manipulator. I worked in so many data-processing institutions and companies that Im now having nightmares about it again. Eventually, I even managed a couple of these centers of simple counting, one for an ad agency and one for an independent market research survey firm.
With unqualified assurance, I can tell you that every time you run any significant universe of data through the same procedure, hand or sorter count, the results will be different. Things will change. They will change in a vacuum. Without any more questionnaires (ballots would be the same thing) coming into or leaving the room, the end result will change every time you run the data. Even the base numbers will change; i.e., how many ballots you think that you started with. We used to joke about it. Look at this, 10 people died, six had babies, 14 got raises, and some even changed sex since we last made a run. No person and no data left the room. No carrier pigeons flew in or out of the window.
There are many reasons why this happens. Its both human and machine error. Of the two, the human error involved in hand tabulating is by far the most changeable. Human error goes all over the place from boredom to misinterpretation to outright fudging. And it doesnt matter if somebody is looking over your shoulder. No one can look into the mind of those making the choices.
The machines are unbiased, without subjective or political preference. And, of course, they are much faster than humans. Yet, even the machines will screw up or do things differently. Cards will stick together, air suction will fail or misplace cards and spit them in different bins, counters will slip and so forth. Chad, or what I think we called chafe, slots or punches, are the least of the problems and the easiest to correct. Just clean the cards before the first run. Im talking about punched holes that are perfect, and you will still have changes every time you recount.
We learned to live with it, just like we lived with the spring-wound watches that had to be cranked up almost daily and gained or lost time with a certain regularity. But no one ever thought that going back to sun dials or sand hourglasses would be better, any more than we thought hand counting would be more accurate than using the machines. Such thought borders on insanity or political motivations.
What did we do about it? Frankly, we usually balanced the books. Remember, the errors were almost always around 1 or 2 percent or less. If we had a big error, we could always find out why with a little effort. When we forced everything to add up to 99 or 100 percent, we did it by throwing the changes into the all others or miscellaneous columns. Sometimes, we dumped them in cells that were already so full that one or two plus or minus votes/choices would hardly change things by more than a 100th of a percent. Other times, we would put them in pre-coded choices few if any had picked because we knew no one would pay much attention anyway, they were already insignificant. This was all SOP. Its like balancing your checkbook once a year or doing the books of a fairly large hotel or business. You havent time or the resources to chase pennies out of balance, just charge it to petty cash of whatever will not hurt anyone. Reconcile and be done with it.
And Ive been giving you the simple side of it, the straight count. What do you suppose it was like when you got into cross-tabulations? What do you suppose happened when you wanted to find out how many 8 year olds bought Cadillacs, and run that against how many mothers spanked their kids?
When you are dealing with political bias, you can be absolutely certain that any changes and corrections will be done in favor of the selectors personal feelings. Joe Baker is absolutely right in pointing out the unfairness of the subjective element applied in a vacuum.
Let me remind you of what we heard from Palm Beach late Saturday night, after midnight. Two a.m. Nov. 12. Forget whos winning, just note the differences each run.
On election night, the first count was 152,862, Bush and 256,945, Gore. The second machine count was 152,954 and 269,696 respectively. The third count was 152,951, Bush and 269,732, Gore. Unless I didnt write them down properly or fast enough, these are the resultant numbers.
And still Mr. Gore will not reconcile.
The real problem
What do you suppose we would have told a client if one of the questions turned out to be split 50-50, particularly if it was a question the client forced us to include, one of his biggest concerns? What conclusions would we draw from a survey of two new products that split dead even? What would you tell a client who paid to survey the entire nation on just one question and more than half of the population refused to be interviewed, to even participate?
Any conscientious report would not be painting pretty pictures. And we would not still be massaging the data in a futile attempt to find differences in remote or atypical parts of the country. Even if these were centers for the super-rich and famous plus all the entourage they require and attract, or vacation and yachting capitals of the world.
And, perhaps most importantly, what would you say if you did happen to cross results by demographics and found that large urban areas behaved dramatically different than the rest of the country? What would you make of that, while you sit in your vacuum?
Would you start pointing to the things that David Reisman said about The Lonely Crowd or what some other social psychologist told us about the animal kingdom only developing homosexuality when kept in cages?
Im sorry, but if it were my client, Id be telling him that youve got much bigger problems than are answered here. Youd better roll up your sleeves and start looking for answers with a better approach to finding out what they are, and resolving them, while you still have the option of choice.