Counting numbers II: Measuring in Regiments, Battalions, Companies
One more point about army sizes: we usually convert army sizes into an absolute number of men, even though numbers of units (regiment or battalion or company) are the most frequently occurring measure in the sources, and sometimes the only size information available. Whenever we are forced to rely on unit totals, we need to constantly remind ourselves of the weaknesses in measuring size this way.
The problems? First, we think in absolute numerical terms, e.g. 5.000 vs. 10.500, and it takes some mental effort or research to figure out what was the full-strength of a regt in period X – as an author we should do this for the reader whenever possible. This is made more mentally taxing when we have to recall the difference between numbers of men in the standard foot and horse units – 30 infantry battalions and 150 cavalry squadrons sounds like a huge disparity in numbers between the two, but aren’t nearly as lopsided in absolute numbers as it seems if squadrons have only 150 troopers. Then, if all you have is a total number of regiments, you’re necessarily comparing Red Delicious apples to crab apples, without knowing how many you have of each. You can (and probably should) list the regiments by name if at all possible, for accuracy and reproducibility of your results. We’ve already discussed some of the issues: regiments might have one or two battalions, depending on the country and type of regiment (e.g. Guards regiments often had two), so you need to make those calculations. Battalions in turn might have increasing or decreasing numbers of companies, especially from peace to wartime. The official strength of the regiment changes over a longer time frame, so you need to correct for that if you are doing any longitudinal comparisons. The sizes of regiments will vary according to the campaign season as well. Weak battalions might have 350 men, whereas a full-strength bttn could be 700, maybe 750 with officers. So with a single bttn you could easily have a range of 350-750 men, or 50% variation. Then, multiply that (making the untenable assumption that all bttns will have the same discount rate) by the number of regts or bttns, and you see how the absolute numbers can be hugely inflated: a difference of 400 men might not be a big deal, but 50 bttns could range anywhere from 17,500 to 37,500 troops – a 20,000 man difference. That’s almost an entire additional field army!
So how do we indicate this margin of error? If we were social scientists, we’d come up with an index I suppose (a strength of thirty-four 1660-bttns). Here’s my suggestion. I think at the least we need to say what type of bttn measurement is being used: men, men+officers, don’t know. This is particularly the case if we’re combining different types of sources. If you took the trouble to do those calculations yourself, let the reader know how you did it so they don’t have to reinvent the wheel. We should also count in bttns rather than regiments, given the possible variation there, even though we need to keep track of the regiments by colonel’s name. We should also give a margin of error. So maybe in 1706 the field army size was 55,000 ± 500 because it was a relatively easy campaign – between an estimated minimum of 54,500 and 55,500 at maximum full-strength. Then in late 1710, after a long series of sieges which saw many casualties and much disease and desertion, the field army size is now 75,500 ± 15,000 (between 60,500 and 90,500). This way the reader gets a sense of how precise those measures are, and how much variation there was from campaign to campaign, perhaps even month to month.
When displaying sizes graphically, the Corvisier example above shows how easy it is to used a stacked bar chart for actual vs. paper strengths. When looking at troops for a particular campaign however, it might make sense to use error bars, or maybe a boxplot if you have multiple possible measures. Just one graph that had each regt on the X axis and # men on the Y axis with error bars for each regt (or aggregated, with campaign year on the X and # men in army on the Y) would really force home the point about variation. I’ll try to make one up sometime when I get a chance, unless someone knows of one already done? Calculating an average discount rate for an army is a start, but even better is to show whether this discount rate is evenly distributed through all the army, or whether there are certain regiments that are way below their strength – you could of course tie this to particular combats. The table below gives us a sense of the wastage rates, but even more of the spotty nature of our sources. Is there a better way to visually display this info?
As the table above suggests, we should also make note of when the sources we’re using were made. Campaigns have a rhythm, so it’s important to know when the numbers were measured during the campaign – at the beginning when the army was at full strength? right before a battle, right after it? at the end of a long exhausting campaign? Sometimes they’ll review the troops because some event just happened (say, a battle) which now makes them doubt their previous figures. Anybody know about the decision process for when to make a review of the full army?
In the past I’ve considered taking precise-sounding numbers (4,325 men) and rounding them off to even-sounding numbers to indicate the imprecision and forestall any sense of spurious accuracy/precision, but this has the disadvantage of making it difficult to trace the sources that you got that info from; at the least, you’d want to wait till the end to round off.
Again, I think some historians have done some, or even most, of these things, but I still see plenty of generic “34,550 men” in published works, when you have no idea where those numbers came from, and how big the real range is likely to be.