Wednesday, August 6, 2008

Playing the percentages--or the raw numbers

In the 2004 Presidential election, George W. Bush received 62,040,610 popular votes; John Kerry gathered 59,028,439. Bush got 50.73% of the popular vote to Kerry's 48.27%. This seems to me to be a case where, if you are discussing the numbers, you'd use the percentages, since it's easier to say "Bush got almost 51% of the popular vote" than to say "Bush received well over sixty-two million votes." Furthermore, it's clear from the first statement that Bush had a slight majority of the popular vote, but it's not clear from the second statement without any data telling how many folks voted for Kerry. (That will matter in 150 years when nobody will know off the top of their head who held the White House from 2005 to 2008.)

Of course, if you live in a small village with thirty voters and seventeen cast their ballots for Bush, you probably wouldn't say that 56.66% of your town chose Bush over Kerry. At that small a number of votes the percentages sound rather arcane; it's much easier to declare "Seventeen voters in our town picked Bush, the other eleven wanted Kerry."

I mention this because when you encounter numbers that are presented as either raw numerals or alternatively as percentages, the person responsible for choosing which format to use is making a judgment on which works better to make the point. But beware: someone can use one format over the other in an attempt to make an argument sound stronger than it is.

I recently encountered a good example of this in the book The Impending Crisis: 1848-1861 by David M. Potter (1976). Let me at the outset admit that I'm making a minor quibble here that in no way diminishes the outstanding nature of the late Professor Potter's book; it's one of the most enjoyable, information packed histories I've ever read.

But no book is perfect, and on pages 505-506 Potter makes this weird point:

"The election contests for (the conventions in upper South states to consider succession) were fought out under circumstances quite different from those of the contests in the lower South. Basically, the upper South was not as obsessively committed to slavery as the lower South. Of the seven states that had (by February 1, 1861) seceded, only Texas had a Negro population of less than 40 percent; in the five states that were about to act (Arkansas, Virginia, Missouri, Tennessee, and North Carolina), the Negro population averaged less than 30 percent." (Emphasis mine).

Okay, let's look at some raw numbers. Here are the actual tallies from the 1860 census on the number of slaves in the slave holding states. This was the last census taken before the Thirteenth Amendment abolished slavery, right at the same time those secessionist conventions were taking place:

State, number of slaves

Alabama 435,080

Arkansas 111,115

Delaware 1,798

Florida 61,745

Georgia 462,198

Kentucky 225,483

Louisiana 331,726

Maryland 87,189

Missouri 114,931

Mississippi 436,631

North Carolina 331,059

South Carolina 402,406

Tennessee 275,719

Texas 182,566

Virginia 490,865

Total: 3,950,511 slaves

The wonder of doing research is that you learn a lot of things, sometimes fun, sometimes sobering. Actually going through the census pages and counting up all the slaves was rather a bitter task; I could picture more readily than I've ever been able to the thousands of people being whipped, having their children taken from them and sold, and being denied any kind of education. And I was struck by the sheer magnitude of those numbers: nearly four million people in bondage. Lincoln in his Second Inaugural Address noted that "One eighth of the whole (1860) population were colored slaves." As I type this, I'm checking the Census Bureau's home page, which says the current population of the United States is 304, 798, 824 people. If one-eighth of us were slaves today, that would be thirty-eight million, ninety-nine thousand, eight hundred fifty-three slaves. That's nearly two million more people than live today in Illinois, Ohio, and Pennsylvania combined.

I should also note parenthetically that if you're surprised the slaves were tallied so diligently, don't be; the Constitution, after all, mandated they be counted to determine the apportionment of members of the House of Representatives. That's the infamous three-fifths clause of Article 1, Section 2; so Alabama's 435,080 slaves counted as 261,048 bodies when they were figuring out how many congressmen the state received.

Anyway, let's go back to the point made by Potter. The upper South was less obsessively committed to slavery than the lower South? Notice that the state having the most slaves, by far, was Virginia, one of those upper South states that didn't secede by February 1861. Not only did the Commonwealth have more slaves than all seven of the lower South states; it had over twice as many as Texas and Florida put together. North Carolina had the second highest number of slaves of any upper South or border state, far more than either Texas or Florida and virtually the same number as Louisiana. Looking at raw numerical counts, it's certainly hard to argue that Virginia and North Carolina were less "obsessed" with slavery than states deeper in Dixie.

Saying that the state that had the most slaves was one of the least obsessed with it is problematic. So how did Potter come to that conclusion? He went by percentages instead of numbers, pitting states with a "Negro population" of over forty percent against those where it was under thirty percent. Let's see how this works out with Virginia, an upper South state and ranking first in total number of slaves, matched with Georgia, a lower South state which ranked second in number of slaves.

Virginia in 1860 had, in addition to 490,865 slaves, 1,105,453 white people. Georgia, besides the 462,198 slaves, had 591,588 white folks. Both states, by the way, had a third category: "free colored people" of which there were 58,042 in Virginia and 3,500 in Georgia. It's easy to forget that the antebellum South had black people who weren't slaves, but it did, which is actually another problem with Potter's statement. He writes "Negro population" when he really means the portion of the Negro population that wasn't free, so he's basically inappropriately combining data sets.

Counting everybody, white, free black, and slave, Virginia had a population of 1,654,360 while Georgia's residents numbered 1,057,286. And yes, a larger percentage of Georgia's residents were slaves, 43.7% of the total as opposed to "only" 29.7% of Virginians held in bondage.

But is this really a case where a percentage is the best way to present the data and make a point? I don't think so; Virginia still had 28,667 more slaves than Georgia; it seems preposterous to me to argue that Virginia was less obsessed with slavery simply on the grounds that it also had half a million more white people.

Besides, the fact that Virginia had more whites than Georgia didn't necessarily have to do with slavery. Some of the difference was that Virginia was becoming more urbanized. I use the term "urbanized" in very relative terms; we are talking about 1860, after all. Still, Virginia had five cities of ten thousand people or more; Georgia had only two such towns. The greatest call for slaves, of course, was for agriculture, which took place outside cities as it does today. The more urban the population, the lower the percentage of slaves among the total population.

Taking all this into account, I'm unpersuaded that Virginia was "less obsessed" with slavery than Georgia. I think using percentages trapped Potter into making a statement that is far from clear.

No comments: