In 1954, Darrell Huff wrote a book called How to Lie with Statistics. I need to read it. And I think some Civil War scholars need to read it, too.
This post is in response to a review by Kenneth Noe at Civil War Monitor. Let me first say that I read Dr. Noe’s book Reluctant Rebels and enjoyed it. It was a solid work of scholarship. Noe is a well respected historian. However, we apparently have different views on the efficacy of using statistical analysis in a work on the Confederate/Civil War soldier.
First, my book is not without statistics. I mention the commonly known statistic that only about 5% of southerners owned slaves in 1860. I also note that this percentage does not include white families in the statistics. Roughly 25% of white families were in the slaveholding class. That percentage rises further when you are discussing families in the Confederate States (for a good breakdown of this, I would consult Richard Bonner’s excellent book, Mastering America).
Dr. Noe, however, is interested mostly in slaveholding as it pertains to Confederate troops. He found it frustrating that I used such vague qualifiers as “some” or “many or most.” More specifically, he didn’t understand why I did not employ a “sample” of soldiers upon which to make conclusions about various topics.
Some Civil War historians (I’m not sure what percentage of the whole) have taken great pride in playing the numbers game. The use of quantitative methods has been a bedrock of the New Social History, which emerged in U.S. scholarship in the 1970s. The use of a “sample” has been the rage in Civil War soldier studies since around the time of James McPherson’s book, For Cause and Comrades, published in 1997. McPherson said things such as “20 percent of soldiers avowed explicit proslavery purposes in their letters and diaries” (110) and that “two-thirds of both Confederate and Union soldiers . . . expressed generalized patriotic motives for fighting.” (114) Now, it seems, every time a scholar writes a book about Civil War soldiers, he/she is expected to say exactly how many he/she obtained data on.
Noe’s frustration over the paucity of statistics in my book is valid, especially given the recent trend in historiography. But is it worth playing this numbers game?
I wrestled with this issue of drawing on a “sample size” for Marching Masters. Previous scholars of the Civil War have made much of having done so, and some of them I discussed in the introduction to Marching Masters. I suppose I could have said that I looked at 500 Confederate soldiers or 1,000 or 750. However, I was never sure how to count them, since any information I drew was not equal in quantity or quality. Does a memoir by one soldier count as much as a single letter by another? Whatever the number I could have looked at–or anyone could have looked at–it is dwarfed by the number I did not, and could not ever possibly have looked at.
Total numbers are sketchy at best, but the CSA army may have included as many as a million troops. They say a statistical sample should be about 10 percent. To obtain a statistical sample of Confederates would mean looking at at least 75,000 and as many as 100,000 troops. No historian has even come close to that. I doubt a historian has even read 100,000 letters (nevermind those of 100,000 soldiers) in order to write a book, which would mean reading 55 letters per day for 5 years.
But soldiers don’t just write letters. They also write diaries, reports, and memoirs. What exactly does it mean to say that you “looked at” a Confederate soldier, anyway? Read a book of his letters, read all the letters in a set of papers in an archive, read his diary? Read every single piece of paper he ever wrote? Sometimes an archival collection may contain only a single item from a soldier. Does looking at such a collection adequately reflect the wartime service of this soldier? No way. So, should then I assume I’ve done all I can do, add one more soldier to my sample size, and then move on?
Historians are constantly making choices about what to leave in, what to take out. But inevitably, the officers get over-represented in Civil War books. Every archive is weighed toward men and women with better educations, financial resources, and writing ability. Conducting historical research is an elitist game in which the thoughts and feelings and rantings of officers, upper-echelon politicians, and newspaper editors far outnumber those of the poorer classes. When it comes to the Civil War, the research pool is not even. Some soldiers wrote a lot, others wrote little. Some got killed in 1861, others lived until 1922. If a soldier had one slave and talked about him, does that make him more important when discussing slavery than one who owned 20 and is silent? How does one quantify that as proslavery feeling?
Some people take comfort in numbers. They seem authoritative. But what are they really telling us? Take Noe’s book Reluctant Rebels. Noe uses numbers to discount certain ideas about Rebel troops. He says that only 29 of his soldiers (9.1 percent) mentioned the word “independence.” (34) The lack of this word he apparently takes as a comment on soldiers’ ideological convictions. But it seems absurd to me to boil down something so political by focusing on key words, especially since there are many ways Confederates could have expressed the idea of independence.
The absence of a word doesn’t suggest the absence of a thought or feeling. Take sex. In the course of all my Civil War research, I don’t recall many details about Confederates’ sex lives. Yet, I can imagine they thought about and engaged in sex. And if I knew a soldier had three children, I could be fairly sure he was not a virgin.
Confederates used trains often. The train was important to the Confederate war effort. But I don’t recall much talk of trains.
The point is, sometimes actions speak more than words–whether it be about patriotism or slavery or sex. A patriotic soldier might talk about “independence,” or “fighting for our rights,” or “freedom,” or “self-rule,” or states rights, or whatever. Or he might not desert his post when some shitheel in his company asks him to run off with him. If someone asked you how patriotic you were, how exactly would you prove it? Would you say that you were patriotic because you paid your taxes or because you refused to on the ground that your taxes supported an unjust war or crooked politicians that you thought were un-American? Would you prove it by waving a flag?
At one point in Reluctant Rebels, Noe discusses Private Grant Taylor, who wrote in 1864 about his fear of subjugation at northern hands. On this subject, Noe uses his soldier sample to score points off of James McPherson, whom, Noe believes, erroneously found the term subjugation “ubiquitous.” In contrast, Noe states, “Only seventeen soldiers in the sample (5.3 percent), roughly representing Lower South and Upper South evenly, joined [Private Grant] Taylor in decrying it.” Where some might say “very few,” Noe would say 5.3%.
And yet, at what point does a number become significant? 5.3% doesn’t seem significant. But what? 10%? 25%? 70%? What are we expecting exactly from these long dead soldiers? And what exactly does quibbling over percentages tell us about the Confederate war effort?
Were I to ask today’s soldiers whether they preferred Coke or Pepsi, I could easily come up with some hard numbers. I might even ask them about whether or not they believed they could eradicate terrorism. But even if I could ask that, what does that tell me about the war effort?
When examining the Confederacy, it doesn’t make sense to take soldiers out of the context of the larger war effort. It’s great to know what percentage of men hated slavery, liked slavery, hated trains, liked trains and sex and pork rations. But the soldier was responsive to decisions made at higher levels. The Confederate army wasn’t simply a mob of men with ideas about certain things. They were subject to taking orders and grand strategy. It didn’t always matter what they thought.
Thus, in Marching Masters I wasn’t obsessed with statistics or looking at the “common soldier” (that is, a statistically normal one–though “normal” is subjective in many respects). I did not weigh privates more heavily than officers, even though privates were far more numerous in the army. My book would have been impossible to write without looking at high-ranking men like Nathan Bedford Forrest and Robert E. Lee. These men gave orders and determined policy. And their ideas changed over time. Lee wielded far more influence than any other man in the army. While it’s nice to know whether or not 53% or 18% or 29% of men in the ranks supported the enlistment of black troops, Lee and Jefferson Davis wanted it, and they got the Confederate Congress to pass a bill to that effect. I think Lee had lots of support in the ranks for this measure. But the army was not a democracy. Why should we be obsessed with what common men thought?
Ideas are important. But in war, actions matter most. A man did not have to be a proslavery theorist to show he was proslavery. He didn’t even have to own slaves.
At a simpler level, Noe complains that I did not include a chart–as he did in his book, Reluctant Rebels–showing how many men I “looked at” from each state, what their age was, number of slaves owned, etc. In Noe’s appendix, he lists how many men he looked at from Alabama. The number is 56.5. How does one get .5 of a person? Noe notes that .5 means a man split his time between two regiments.
However, one could ask, did the soldier really split his time between these two regiments equally? Or maybe 40% between one and 60% between another? 75/25? One could question the validity of splitting the difference.
At a more serious level, Noe’s sample of soldiers is heavily weighed toward Alabamans, who make up 17.7% of his sample, despite the fact that Alabamans made up only 9% of all Confederate soldiers. Why are Alabamans so prominent in Noe’s book? I reckon it’s because Noe is a professor at Auburn University, which would have a lot more resources about Alabama than say, Rice or Vanderbilt or UVA.
In his sample, Noe included one soldier from Kentucky, one from Maryland, and 2 from Missouri. The number of men from Texas in his sample equals the number of men in his sample from Arkansas, Florida, Kentucky, Missouri, and Maryland combined. Can one soldier from Kentucky–never mind Missouri–actually speak for the experiences and attitudes of thousands of men from those states? No.
And regardless of what this one soldier from Kentucky (note: Abraham Lincoln, Jefferson Davis, John Bell Hood, Albert S. Johnston, and Braxton Bragg were all born in Kentucky) thought or did, it likely had little impact on the war as a whole. Kentucky was far more important than what any one soldier said about it. In my book, I did look at more than one Kentuckian. But what was more important was the importance of Kentucky as a slaveholding state (Kentucky had more slaves than any state outside the Confederacy) and the efforts by the Confederate army to conquer it.
One can find other misrepresentations in the number of slaveholders in Noe’s sample. Of his total of 320, there were 92 men who owned slaves. This is 29%. Only 5% of southerners owned slaves in 1860. This percentage increases if one looks at only future Confederate States. But it does not come close to approaching 29%.
And yet, Noe’s numbers reflect a larger reality: slaveholders were over-represented in the Confederate army. When it comes to slaveholding, one simply cannot look at any large number of Confederate soldiers that will reflect the slaveholding demographics of the South as a whole.
Among Noe’s slaveholders, 21% would have been considered planters, who owned 20 or more slaves, even though planters made up less than 1% of the slaveholding classes. Another 21% owned 10-19 slaves. The average slaveholder in the Old South owned 5 slaves. The largest number of slaveholders in the Old South consisted of those who owned one slave. Yet, only 8 in Noe’s sample owned one. 64% percent owned more than 5. All of these numbers are wildly unrepresentative.
But does it matter? Does the fact that Noe does not contain a representative number of the South’s slaveholders invalidate his claims? Certainly not. But it does call into question the importance of this kind of number crunching, especially since Noe and I do not disagree about the general attitudes among Confederates. Our difference is mostly a question of emphasis. I put slavery at the heart of the Confederate war effort, while Noe does not.
I think quantitative methods is great if you are looking at a sample of people about which you can gain easily measurable data. if I keep polling people about whether they like Coke or Pepsi better and I inevitably get a pro-Coke result, I can probably conclude that most people prefer Coke.
But no one ever took a poll of Confederate soldiers to see whether or not they thought they were fighting a war for slavery. And even then, the results might have changed based on where and when you asked your questions. Regardless, need we always believe the results of a poll? Polls themselves can be flawed.
Why should a military historian be held to using quantitative methods anymore than someone writing about any aspect of slavery? Or anything? Would someone who had written a book about feminism be called to task for not asserting what percentage of feminists thought about a certain idea? Do feminists historians routinely say their research included a “sample” of women? I think not.
In my introduction, I admit that my book is admittedly impressionistic in nature. But so was Bell Wiley’s classic book Johnny Reb, a 70 year old work that is still in print. I think statistics are good when you want to quantify something concrete: how many troops served in combat, how many were casualties, how many deserted. But these numbers are hard to come by. Quantifying something as inherently subjective as racism or patriotism is even more difficult for the historian. And perhaps, something not worth pursuing.