Problems with Pensions

This is the first in a series of posts defending the view that a relatively small number of Americans fought at Cowpens.

Previously, I reviewed primary sources, and to a lesser extent, secondary sources, and arrived at a rough estimate of the number of British and American participants at the battle of Cowpens. Specifically, I suggested that the British had around 1,150 men, and the Americans had somewhere between 800 and 980 men.

Of note is that Lawrence Babits in his recent and influential history of the battle, A Devil of a Whipping: The Battle of Cowpens, arrived at a very different determination, for at least the American army. Babits concluded that Morgan may in fact have had as many as 2,400 men under him; or three times the number of men Morgan claimed in his after action report.

Unlike a number of earlier histories, such as James Graham's The Life of General Daniel Morgan... (1856), which I brought up before, Babits explained the sources and reasoning behind his totals. His approach both deserves and allows careful consideration.

Babits estimated American numbers using several different approaches. For the number of men in the commands of Captain Samuel Hammond of South Carolina and Major John Cunningham of Georgia, Babits deferred to past authority (he cited a study described in a footnote in the Greene Papers). However, for other units serving on the militia and skirmish lines, namely the commands of Colonel Thomas Brandon, Colonel Joseph Hayes, Colonel John Thomas, Lieutenant-Colonel Benjamin Roebuck, and Major Joseph McDowell, he relied on estimates derived from pension records. His method of using pension records to determine unit size is the focus of this post.

Babits' method of using pension records to estimate unit size is innovative and commendable. His method can be summarized as follows. First, one must identify units where the number of participants is known. Second, one must identify the number of pension applications generated by individuals serving in those units. Third, one must calculate the ratio of pension applications to participants for those units. Fourth and finally, one then estimates the size of other units by extrapolation. That is, the number of pension applications generated by other units is multiplied by the pensioner-to-participant ratio.

So long as the "known" unit sizes are accurate, and the probability that a participant of the battle would have generated a surviving pension application is constant across units, then there are no serious issues with the validity of Babits' method. (Although the estimate would have a smaller margin of error for larger units than smaller ones). Problems begin to arise if the men in some units were less likely to survive into old age and file a pension application than men in other units, or if pension applications were lost or destroyed in subsequent years in a nonrandom manner.

It stands to reason, that the problems I alluded to above would have occurred in some cases. My own cursory inspection of pension records uncovered surprisingly few from Georgians and South Carolinians serving with Major John Cunningham and Captain Samuel Hammond. If Babits had the same experience, then this would perhaps explain why he relied on past authority, and not his own estimate, in determining the number of men for those units. For related reasons, it makes sense to draw a distinction between units of regulars and units of militia as the two bodies of men tended to be drawn from different segments of society and faced differing degrees of hazards during the war. The regulars, presumably, would have been less likely to survive into old age.

In A Devil of a Whipping, Babits identifies several militia units at Cowpens with a known unit size. These are Captain James Gilmore's company of Virginia militiamen (38), Captain Mordecai Clark's company of North Carolinians (25), Captain Samuel Sexton's company of South Carolinians (24), Captain Samuel Otterson's company of South Carolinians (30), Captain John Collins' company of South Carolinians (24), and Captain John Irby's company of South Carolinians (60-70).

There is, unfortunately, a problem with these numbers that is independent of any problems with the method. Babits shows 38 men in Gilmore's company on page 31 and 44 in Table 1 on p 32. For my calculations below I used the smaller of the two numbers. Babits has 20-30 men for Clark's company and 60-70 men for Irby's company, I used the midpoint of those ranges. I haven't checked Babits' numbers for Gilmore and Clark, but I have for Sexton, Otterson, Collins, and Irby. With the exception of Irby, the numbers for these companies may be inaccurate.

Sexton, or rather, Saxton, said in his pension application (pdf) that, "I succeeded in inducing twenty-five men to join me, and was chosen their captain;" this would be a company of 26 men.

Otterson's 30 men were not of a company he commanded, nor did this group fight at the battle. Otterson recalled (pdf), "This applicant states that he omitted to name that he & the regiment commanded by Brandon to which he belonged was under the command of Col. Morgan at the Battle of the Cowpens but that he with several others about thirty were sent out as spies some days before the engagement at the Cowpens & from some cause did not arrive until the Battle was over."

Collins said in his application (pdf), "on the night before the Battle of the Cowpens I again joined General Morgan with 24 fresh men, and fought with my company the next day." This would be a company of 25 men.

Irby's statement can be found here (pdf).

Without making adjustments, Babits' numbers for these militia units include a combined 206 men in the units and 18 pension applications by survivors. This a pension-to-participant ratio of 1:11.4. Now the process of extrapolation begins. Research uncovered a total of 108 pension applications for North Carolinians at the battle (Babits, p 37). Babits noted that one cannot readily tell in many cases which of these were from veterans in Major Joseph McDowell's battalion of North Carolina militia, Captain Henry Connelly's company of North Carolina state troops, or from North Carolina Continental infantry or cavalry. It is safe to say, however, that most of these veterans would have been with McDowell, whose battalion was the only large body of North Carolinians at the battle. Taking the conservative estimate that 75% of the 108 North Carolina pensions were submitted by men that had been in McDowell's command, then it would be projected that there were 923 men in McDowell's command (81 pension applications * 11.4 men per application).

The projected total is likely to be greatly in error. Brigadier-General Daniel Morgan wrote to Major-General Nathanael Greene in late December that, "General Davidson has brought in one hundred and twenty men;" much to Morgan's disappointment Davidson did not furnish additional reinforcements. This force of 120 men was probably McDowell's command at the battle; the number is less than 1/7 the men predicted by the pension method.

How can the pension method provide an estimate that is so far off the mark? As I said before, the logic behind the method is sound. The problem, I think, is that the initial assumptions about unit size are invalid. To take one example, the Captain Irby mentioned above claimed that a band of 60-70 men elected him captain before the start of the battle. This command, in other words, was at least half the size of McDowell's command according to Morgan. Only the pension applications of Irby and one other soldier identify Irby as a company commander at Cowpens. No other account of the battle mentions his company. It is therefore reasonable to conclude that Irby did not actually command a force of 60-70 men at Cowpens. Actually the pension application does not quite say that he did. The relevant excerpt from his pension application is as follows:

"he [Irby] with others fled to the Army of Morgan's Army for protection some time before the Battle of the Cowpens – that there were many refugees then in a similar situation with himself – that they formed a volunteer Company to the amount of 60 or 70 and that he was elected Captain of said Company and was commissioned as such by General Pickens of South Carolina – that he served as Captain of said Company in the Battle of the Cowpens – at which Place Tarleton was defeated."

Irby's election to the captaincy occurred "some time before the Battle" and for all one can tell some of the 60-70 men that elected him might have decided to go back home or join up with other units before the battle occurred. Then there is also the fact that this statement appeared in a pension application written 52 years after the battle. Was Irby's recollection accurate? Might not he have overstated his importance during the battle? After all, he was trying to convince the government to award him a pension. Similar concerns can be raised whenever the "known" unit size depends on a statement in a pension application.

To be fair, Babits did not quite take the approach that I've described here. His pension-to-participant estimates were based on combined records for regulars and militia. (Counterintuitively, the regulars have a lower pensioner-to-participant ratio than the militia). He also seems to have regarded Irby as an outlier (although he didn't cast doubt on the veracity of Irby's statement). If one adds up the numbers of participants and the numbers of pension applicants for these units (listed in Babits' Table 1, on page 32), one finds a total of 88 pension applications representing approximately 447 men, or 1 application for every 5.1 participants. When this smaller ratio is applied to the North Carolina numbers, the calculated estimate is 413 men for McDowell's battalion. This estimate is still several times larger than the number indicated by Morgan. Babits himself only loosely followed the method he described. He stated that the pensioner-to-participant was about 3 or 4 to 1 (despite the numbers he was using), and he estimated McDowell's battalion as having only 200+ men in it. It is unclear why he estimated McDowell as having so few men. Again assuming that only 75% of the North Carolina applications were from men with McDowell, then the pensioner-to-participant ratio would have had to have been a mere 1:2.47 to arrive at this total.

A counterargument to the position I've adopted would be that it seems unlikely that 81 of the 120 men in McDowell's command would have lived long enough to have been able to file a pension application. I'll concede the point. It doesn't follow, however, that this makes the pension-derived estimates more credible and not less. I will address the issue of pensions and veteran survival in my next post.

Related: The Battle of Cowpens in Miniature, Part 1, How Many Fought at Cowpens?, The Militia Line: Composition and Organization