One of the first questions anyone asks about past societies is, ‘how many people were there?’?It is one of those basic bits of information and generally speaking those of us who teach these societies are usually prepared with an...
One of the first questions anyone asks about past societies is, ‘how many people were there?’?It is one of those basic bits of information and generally speaking those of us who teach these societies are usually prepared with an estimate to answer the question.?But what we often don’t talk about is how we come to those estimates or how reliable they are and that can lead to an excessive amount of confidence in estimates that are often deeply uncertain – especially when those estimates get used by other disciplines and scholars who seem unaware or uncaring as to how thin their evidentiary support is.
So this week, we’re going to look at how we go about answering this question, which is to say the various methods we can use to get a sense of ancient population figures, along with their drawbacks.?I am, of course, going to be particularly drawing on methods used for the ancient Mediterranean, because that’s what I know best.?So we’re going to look at some of the evidence which sits behind things like those neat little block graphs I’ve used in the past.
But I can lead with the upshot here: despite the apparent confidence of works purporting to present historical populations over broad periods, we really don’t know the population of most of these places at most of these times.?In particular, venerable books like the McEvedy and Jones, Atlas of World Population History(1978) are mostly built on sand and yet they in turn get used as the foundation for all sorts of social science work by social scientists who assume the confident conclusions presented in these sorts of books are, you know, confident and who lack the area-specific skillset to assess them.1
As always, if you like what you are reading here, please share it; if you really like it, you can support me on Patreon! If you want updates whenever a new post appears, you can click below for email updates or follow me on twitter (@BretDevereaux) for updates as to new posts as well as my occasional ancient history, foreign policy or military history musings, assuming there is still a Twitter by the time this post goes live. I am also on Bluesky (@bretdevereaux.bsky.social) and (less frequently) Mastodon (@bretdevereaux@historians.social).
Population Figures in the Sources
Understandably just about the first place anyone starts with this question is to check and see if the sources just tell you how many people there are.
Immediately, however, we have to divide these kinds of reports into two categories, divided by a very basic question, “did they actually count?”?That is to say, is this population figure the clear product of an effort to conduct some kind of a census (and thus perhaps reliable) or is it a blind-guess figure or worse yet a symbolic figure, a number chosen to mean “a lot” or “not many.”?That last possibility, figures chosen for their symbolic importance, may seem strange, but it is remarkably common in ancient literature. If that seems strange, well, think of how often in regular speech people describe something big as ‘a ton’ or ‘a million’ not to mean an exact figure, but just ‘a lot.’?Our sources often use stock numbers (like a ‘myriad’ which is notionally 10,000) to mean ‘a lot.’
The gold standard for reported figures is, of course, the Roman census.?In that case, we know that the Romans did in fact count and that our authors reporting those figures are doing so with access to the official records.?Moreover, we know that the documents produced didn’t just render a number, but rather a list of citizen-households, with their members, age and property, which makes an accurate count more likely; that relative accuracy was a product of the census being essential to the conduct of Roman taxation in Italy (before 167 when tributum ended in Italy) and Roman conscription.?So while those census figures aren’t perfect (historians in particular debate what percentage of the population might have been missed), they’re pretty damn good.
Detail from the so-called Altar of Domitius Ahenobarbus.?Just off of this image to the right is a scene of a lustratio, a Roman purification ritual associated with the census.?Consequently, the seated man on the left has been interpreted as one of the censors, recording a census, perhaps as part of the foundation of a new Roman colony.?On this monument, see F. Stilp, Mariage et Suovetaurilia: étude sur le soi-disant ” Autel de Domitius Ahenobarbus “ (2001); alas the work is only in French but is the thing to read.
Of course that immediately just moves the debate from the veracity of the count to the question of what is being counted.?The bulk of our key census reports are delivered by Polybius, Livy and Augustus himself (in the Res Gestae), covering different years.?Livy and Polybius are generally relatively clear about who is being counted.?Livy’s standard formula is censa sunt capita civium tot, ‘so many heads of citizens were counted’ (implying, neatly, that he is including poor citizens, the capite censi, ‘those counted by their heads’), while Polybius’ figure for 225 is a military one, composed of men of military age liable for conscription, citizen and non.?What becomes tricky is the Augustan figures.?The bulk of scholars generally assume that Augustus deviated from the practice of the Republic of only reporting the total count of men and has counted women and children as well as men, because that produces a population growth curve which makes the most sense and follows archaeological settlement data; this is called the ‘low count.’2?Arguing that Augustus has kept to the old practices would produce a total population would produce a population in Italy under Augustus more than double that generally accepted by scholars.3?And then finally there is a ‘middle’ count, which assumes that what is actually being counted were individuals who are sui iuris (legally independent), which after a few twists and turns, produces a figure only modestly higher than the ‘low count.’4
All of which is to say that even when we have an official census we generally figure is quite reliable and which has been faithfully transmitted to us, there is ample ground for uncertainty and confusion, though I must note that I think the debate about the population of Roman Italy has largely settled and the standard range of 4-5 million free persons (note that none of these figures include enslaved persons!) in peninsular Italy (to include Cisalpine Gaul) during the republic, rising to c. 6m by the early empire, is fairly secure.
A diagram of the social classes and population of Roman Italy following Polybius’ figures (2.24).?This is a ‘low count’ model of the population and not quite perfect; at some point I want to go back and revise it somewhat to fit De Ligt’s reading, op. cit. of Polybius’ figures.?That would actually somewhat increase the free population and somewhat decrease the enslaved population.?On the interpretation of Polybius’ figures, see Taylor, Soldiers & Silver: Mobilizing Resources in the Age of Roman Conquest (2020)
The moment we move beyond Roman census data for Italy (while the Romans did conduct census (the Latin plural of census is census) in the provinces, the figures from these do not survive) reported population statistics get terribly murky, terribly quickly.?These figures are often presented without any sense of how they were reached and frequently seem implausible if not impossible when assessed by the other methods we’re going to discuss here.
So for instance, Diodorus (1.31.6-8) says 7 million people lived in Egypt (first cent. BC); Josephus (BJ 2.385) says it was 7.5 million, excluding Alexandria (first cent. AD), that point about Alexandria’s exclusion being important because it was a massive city that would likely add at least a couple hundred thousand more people.?Those figures seem possible, but more than a bit high and scholars have tended to discount them,5 with a range of estimates from the Ptolemaic period to the Roman one clustering between 4 and 7 million.
On the more absurd end we have some city population figures.?Strabo (17.3.15) reports that before the Third Punic War ‘seventy myriads’ (700,000) people lived in the city of Carthage, a figure so high few scholars would accept an estimate even half of that.6?Carthage was a big city, to be sure, but 700,000 at that early date would have been absolutely massive for the western Mediterranean; Rome itself would only get that large in the late first century.
And then there is the hot mess that is the sources on Athens.?You might imagine that the best attested polis in the whole of the ancient world would have some reliable population figures, but what we have is messy.?Thucydides (2.13) gives an overview of Athens’ military strength; the difficulties with that approach are discussed below.?Meanwhile, Athenaeus (6 272cd) quotes a census reported by another source (Ctesicles) taken by another figure (Demetrius of Phaleron) in the late fourth century which reported 21,000 citizens, 10,000 metics and 400,000 slaves.?The problem with that is of course that while the first two figures may be reasonably plausible (c. 20-30,000 citizen males is generally what we suppose), the last is not – it would make Attica (the territory of Athens) the most densely populated human terrain in the pre-modern Mediterranean or quite possibly anywhere, which it very clearly was not.?Alain Bresson concludes the figure is metaphorical and he is, I think, clearly right.7?
Block diagram of what seems to be the rough current consensus position on the demographic structure of Athens by social class, though there is, as noted, a lot of uncertainty here.?The number of enslaved people is, in particular, deeply conjectural and this reconstruction was made with A. Bresson, op. cit., c. 40% estimate in mind, but it may be that there are somewhat more free persons in Athens and perhaps somewhat fewer non-free.
As you can see, once we get past the Roman census figures with their uncommon reliability, the evidence gets remarkably poor, remarkably quickly.?The figures we are given in particular tend to be too high, sometimes ‘plausible, but on the high end’ but frequently absurdly high.?Of course the consequence of that is that, notwithstanding the odd Roman exception, nearly all of the ancient population figures we have are doubtful, at the very least to the degree that they would require some kind of external confirmation before becoming useful, where they are not wholly useless in assessing population (though, of course, still useful in assessing the perception of population).
So if ancient reported populations mostly can’t help, how do we go about estimating, even in a very general, error-prone sort of way, how many people lived in a given place?
Counting My Army Men
The first recourse is to look for another sort of population-related figure out sources might report which would be reliable, and here the recourse is to reported army sizes.?While ancient sources rarely report total population figures and even rarer still would be likely to actually know, they report the size of armies very frequently, as war is one of those activities which deeply interests our aristocratic ancient writers.?This was the core approach of Julius Beloch in Die Bevölkerung Der Griechisch-Roemischen Welt (1886) for the Greek poleis, attempting to derive the number of people from the number of hoplites each polis deployed in its largest deployment, assuming that this figure would basically match the size of the adult male citizen body with minor adjustments and that the rest of the population could be computed from there.?And subsequent efforts to compute the population of classical Greece have often offered some significant deference to Beloch’s figures (often, I’d argue, excessive deference, to the point of not fully following through on the conclusions their own approach implied).
But you may already be coming up with the considerable problems with this approach and there are many.?Of course at the get go one must question if the reported sizes of these armies are reliable.?For many periods and places, that would be an insurmountable objection, but we have good reason to suppose this was tracked by Greek poleis (Athens, at least, did keep a registry of citizen males): accurate figures were both possible and desirable.?And for the most part, the figures we get for the size of hoplite polis armies are not unreasonable, so we generally assume they are more or less accurate.
The next problem is more critical: this assumes that the great majority of the citizen body is in the hoplite class – that is, that there are very few people in this society who are free citizens, but too poor to serve in the army.?Except this is a substantial assumption.?Hoplite equipment in Italy was associated not merely with landed farmers, but with a relative elite – in the Servian Constitution (itself an anachronistic creation of our sources and in Italy, but bear with me) only the First Class of the Roman infantry were expected to serve as hoplites (Livy 1.43; Dionysius Ant. Rom. 4.16) out of five classes of infantry.?The one polis we can see most clearly – Athens – seems to have a pretty significant sub-hoplite class given the economic role that jury pay and enrollment in the fleet and the political importance of their enfranchisement (and our sources apparent assumption that, once enfranchised, they – as a class – dominated the assembly).?Beloch’s assumption is that Athens is unusual in this regard – and that’s not an entirely insane supposition – and that the ‘typical’ polis would have consisted almost entirely of freeholding farmers – a far more courageous assumption.
Immediately following on that problem is the simple fact that militarization rates are highly variable: that is, looking more broadly at pre-modern societies, some put a very high proportion of their military-aged males in service and some do not.?Michael Taylor’s work on the manpower deployments of the second century Mediterranean great powers (in Soldiers & Silver (2020)) demonstrates very clearly how untethered ‘manpower effective’ – that is, the actual number of troops deployed – can be from total population sizes.?To make this problem worse, the one ancient society where we can see this dynamic clearly is also very clearly the most unusual: Rome.?Rome’s militarization rates are extremely and deeply unusually high (how this is possible forms the core question of my book project).?Assuming, as Beloch does, that other ancient city-states must behave rather like Rome and thus have mobilizations which include all or nearly all of their military-aged citizen males, seems unwise.
Attempting to use this method beyond relatively small city-states and instead trying to apply it to large empires thus runs into immediately insurmountable problems, because the ‘men not counted’ (that is, free males not serving in the military) suddenly surge from being a possibly significant minority of the population to a massive super-majority of the population.?The Seleucid Empire, to take one example, seems to have had a maximum military capacity around 80,000 men; being generous we can assume maybe another 20,000 in garrison duty (though in practice, many of these garrisons seem thinned out for major military activity, so we’re at risk of double-counting here), giving us perhaps 100,000 troops total, out of an empire that may have numbered anywhere from 10-20 million – so roughly 2.5-5m military aged males; the effective military manpower is anywhere from 4-8% of the free, adult male population.?In contrast the figure for the contemporary Roman Republic that, in a single year approaches 25% and for Roman citizens over the first four years of the Second Punic War seems to have been above 70%; the possibility for variation here is enormous – so much depends on where on that sliding scale from 4% to 25% you think the society you are looking at is.
All of which intensifies the core problem of counting the uncounted in these figures.?Figures of these sorts aren’t counting women, children or the very elderly, so even if they did scoop up every single military-aged male, we’d expect them to miss about three-quarters of the population (pre-modern societies are very child-heavy to make up for very high infant and child-mortality, so at any given time, probably close to half of these societies are under military age).?And then we have the problems of non-citizens (do they serve? are they counted?) and the non-free (that is, enslaved persons) who are almost invariably not counted and do not serve in the military.
In the case of the Roman figures, the actual census figures put us on a firm enough basis to pin down a reasonably range for these other figures: poor-but-free persons are captured in the census numbers (particularly in Polybius’ figures for 225 which include the socii) and with a relatively complete grasp of the free population we can look to the other methods we’re about to discuss (urban density, arable land) to get a good sense of how much ‘space’ is available for the enslaved population and thus very roughly its size (an assumption we can then test against the sources to see if it is plausible).?But the smaller and smaller one makes the counted population and the less clear its boundaries become, the more the question-marks overwhelm the information you have.
And so, for basically any place that isn’t Roman Italy, the question-marks do overwhelm the information, leading to the search for alternative approaches.
Arable Land, Urban Spaces and Agricultural Modeling
Most of these alternative approaches are focused on trying to take the sort of evidence we can glean and then trying to use that to model or compute the population from them.?The most basic of these is the carrying capacity study: how many people, at a reasonable rate of agricultural fertility, could the land support??The problem of course is that just because a population could exist doesn’t mean that it did exist; few pre-modern populations actually approached the true maximum carrying capacity of their regions.?As we talked about with agriculture, these sorts of societies often end up ‘stuck’ at a ‘low equilibrium’ for extended periods and reaching agricultural carrying capacity would require pushing up to a ‘high equilibrium’ and staying there for some time as marginal land was cleared and brought into production to account for the rising population.
The more sophisticated approach to this we might then term an arable land and population density approach.?In ancient Greece, the classic example of this approach is Corvisier and Suder, La population de l’Antitiquité classique (2000); the authors divide Greece into zones of population density and then compute the total population of Greece.?But this is a method that is clearly entirely dependent on its assumptions: all depends on the population densities chosen.?On the one hand, Corvisier and Suder’s own estimates produce a figure, c. 3m inhabitants in Greece, including Macedonia, Epirus and Thessaly, which is suspiciously close to Beloch’s estimates; one feels, I will admit, a bit of concern that the convergence was intentional.?On the other hand, that figure also doesn’t seem right; the general consensus, so far as I can tell, is that Beloch is too low for Greece for the reasons stated above – he hasn’t accounted for enough non-hoplites be they free-poor or enslaved – so matching his figures is not a good sign.?This is a method that is better than blind guessing, but bound to be profoundly imprecise, offering order-of-magnitude at best estimates.
A better approach is to try and use settlement data, drawn either from archaeology or from literary evidence.?In Greece, this was the approach of M.H. Hansen, The shogtun method: the demography and ancient Greek city-state culture (2006). Hansen takes advantage of the already existing M.H. Hansen and T.H. Nielsen, An Inventory of Archaic and Classical Poleis (2004) which provided a count of every known polis with a rough size-classification for each one.?He could then create a population range for each set of size classifications, creating a range for the total population of the poleis, which Hansen computes at around 4-6m for Greece proper (and as many as 10m for every polis across the Mediterranean).?This approach is also chained to its assumptions, namely the population ranges for small, medium and large poleis, but given the nature of our sources, that’s something we might have a clearer grasp on than population density.?In particular, Hansen tries to get a grasp on those figures by focusing on cities which have been excavated archaeologically and using their area of urban settlement – the built up area of the city – to compute a total population based on urban population densities.?That is a bit more grounded of an approach than arable land because you are assessing visible settlements rather than empty land which may – or may not! – have been settled.
Which is why it is so striking that Hansen’s estimate is substantially higher, to the point of being double what Beloch or Corvisier and Suder (inter alia) produced.?Hansen’s method suggests there were quite a lot more Greeks than we supposed before.?His conclusions have not, so far as I can tell, gained widespread adoption, but for my part I suspect he is probably right or at least closer to right because I think a lot of the other methods have imported assumptions from the structures of Roman Italy – a population absolutely dominated by the freeholding assidui (households liable for military service) – which I don’t think holds for Greece where it seems like there were more enslaved people and the hoplite class was more exclusive.?But the confidence we can place on those assumptions is very low and in the meantime we’re left with a range of populations from 3 million to 6 million – quite the range!
While we’re here, McEvedy and Jones8 suggest a population for the area on the modern borders of Greece of 2.5 million in 400 BC falling to 2 million by AD 1, which is conspicuously not in any of these ranges.?
Similar methods for Roman Italy can – no surprise here – produce much better estimates, because archaeological data can be married to the existing census figures as a means of interpreting them.?This is the approach notably of Launaro, Peasants and Slaves: The Rural Population of Roman Italy (2011) and De Ligt, Peasants, Citizens and Soldiers (2012).?They have the advantage of working in Italy, which is probably the most extensively excavated region on Earth.?We’ve talked about the import of archaeological field surveys which track settlement patterns over larger areas rather than exhaustively excavating individual settlements; once you get enough good field surveys (and you need a lot of them) you can start drawing overall conclusions about settlement patterns – things like “is new land coming into or leaving cultivation?” “Are new villages forming, or are old villages vanishing?” “Are urban centers growing or shrinking – and how large are they and how many people might fit in them?”?When you have some census figures, even very approximate answers to these questions can clarify how you understand those census figures.
Of course once we move out of Greece and Italy things get more difficult and we’re often back to squinting at weaker archaeological data and arable land studies relying on plausible population density figures.?Making matters worse, we can be quite sure that ‘barbarian’ Europe was well below its ‘carrying capacity’ (much land now under cultivation was then forested, for instance) and the army size numbers we get reported for various groups are being reported by their enemies (the Romans, mostly) and are thus also less reliable.?Consequently, the sort of regional estimates one gets for places like pre-Roman Gaul or Britain are really order-of-magnitude estimates and senses of relative proportion.?So long as they are understood like that they can be valuable!?It is certainly the case that we do not know how many people lived in the Roman Empire in 14AD, but it is also true that we can know that it was, say, not so low as 10m, nor so high as 100m.?’Around 50m’ is, by the by, the generally accepted figure, but I wouldn’t be surprised to see it revised significantly as the years go by (more likely upwards, in my view, than downwards).9
Counting the Uncounted
These modeling-reliant methods are also handy, when you have a firm number, for estimating the uncounted.?Note how nearly all of our figures only count military-aged males or male citizens.?That leaves a number of different groups uncounted: women, children and enslaved persons in effectively all cases.?But also sometimes men who are too old to fight or too poor to fight.?How do we count them?
Via Wikipedia, a fresco showing market activity, with merchants showing off wares of fabric (left) and goods in pots (center) from the House of Julia Felix at Pompeii, first century CE.?Note the mix of men and women, adults and children.?Ancient societies were more than just men!
We generally estimate uncounted women, children and elderly males by using demographic modeling based on model life tables – data tables which project mortality simulations based on real-world populations.?But how do you know what tables to use??The answer here has typically been that you comb the evidence you have (grave inscriptions, fragmentary census records that survive in Egypt) to create a statistical snapshot of your population and then try to match that, as best you can, to one of the extant models that ‘best fits.’?That’s tricky too of course: your sources are bound to under-report certain groups (especially the very young), and to give dates and ages which are, at best, approximate.?For the Roman world, the general consensus is that Model West Level 3 (or close to that) from Coale and Demeny, Regional Model Life Tables and Stable Populations (1983) is the ‘best fit’ for the data we have.10?My own view is that you have to use what you have, and these life tables are what we have, but that some of Walter Scheidel’s arguments are well-taken that our ‘error bars’ may be rather larger than we think.
That leaves the question of how many enslaved persons and those too poor to be counted in some of our figures.
Once again in Roman Italy our situation is perhaps a touch better, although not particularly good.?Because we have better demographic data for the rest of the free population and a better grasp on the mechanics of Roman agriculture, we can do something a bit better than blind guessing here (though blind guessing there has been in abundance), but not much better.?Walter Scheidel walked through much of this math, guesstimating the urban enslaved population under Augustus (r. 31BC-14AD) at around 600,000 (based on an estimated breakdown of Roman social classes and estimates of how many enslaved persons each elite household might have) and another c. 600,000 in the countryside (based on rather more confident agricultural modeling showing the figure can’t be much higher than this without pushing out all of the small farmers and tenants we know from our sources there were) for a total of 1.2m probably representing the height of the enslaved population in Italy, coming as it does at the explosive conclusion of Rome’s long streak of rapid expansion in warfare.11?For an Italy under August between perhaps 5.7m and 7.0m that would imply an enslaved population of very roughly 15-20%, with a bit of wiggle room on both sides.12?This, we can be quite sure, is a significant increase from the earlier period so the figure for 225 BC and the Middle Republic must be lower, perhaps very roughly around 10%.?Again, those figures are very rough, but I think Scheidel13 does a good job showing that something much higher, say, 30+% simply doesn’t make much sense given our evidence.
That actually is a useful conclusion, by the by, the full import of which I think has not been fully observed: while Rome has a reputation as the slaveholding society – and it certainly was a slaveholding society, make no mistake – those rates are probably rather lower than what we see in Greece, suggestive of an Italy countryside in particular that had more freeholding citizen small farmers than a comparable Greek polis.?Comparisons with the Near East are more fraught; Egypt seems to have been modestly less enslaved than Roman Italy in the Roman period, but reaching earlier or elsewhere is difficult.
By contrast, as I’ve noted before, estimates for the number of enslaved helots in Sparta range wildly from 75,000 to somewhat north of 200,000.?Agricultural modeling and arable land approaches, used by Figueira, establish that lower bound, while the upper-bound is the product of extrapolating from accepting the 1:7 ratio implied by Herodotus’ report of the size of the helot force accompanying the Spartan army at the Battle of Plataea (Hdt. 9.28.2).?The lower-bound is wedged against the lowest feasible figure from an arable land perspective, but the upper-bound is plausible; I tend to think Hodkinson’s split-the-difference guesstimate of 162,000 is, if not right, reasonably plausible.14
My block-chart diagram of Spartan population by social class, assuming c. 200,000 helots.?On the twists and turns of this argument, see the series retrospective.
But then what of the uncounted in the rest of Greece??These simple answer is we don’t know.?Helotry and the economic system in Sparta was odd and so it attracts attention from our sources and their description of the system lets us chart it; it helps that there is no labor interchangeability between helots and poor spartiates, because there are no poor spartiates.?But in the rest of Greece, you have no merely large, leisured landholders like the spartiates, but smaller farmers – sometimes very small and very poor – plus landless free persons, both citizens and non-citizens and also a large enslaved population.?So the key question – what slice of the poor rural and urban classes were enslaved – is awfully difficult to tease out, with frustratingly limited evidence.?As noted above, ancient figures for the number of enslaved persons are frequently metaphorical big numbers that mean ‘a lot,’ yet at the same time, it is clear the number was ‘a lot.’?The figure ‘one third’ gets thrown around a lot, based on the assumption that the number of enslaved was basically equal to the citizenry and/or that, because this is roughly the figure for the American South before the Civil War, it is reasonable to extrapolate it to Greece (it is not).15?Higher estimates exist; Alain Bresson figures 40-50%, but presents little evidence to support this16 while Nemanja Vujcic in a review of the evidence recently argues for a figure between 15 and 30%.
In fact, what we may mostly be able to say here is that the average Greek polis was both meaningfully less enslaved that Sparta, but meaningfully more enslaved than Roman Italy.?Further precision is difficult.?Once again, moving westward, the evidence vanishes into insignificance rapidly; it seems clear there was slavery in pre-Roman Gaul, Spain and Britain, but now how much.?It seems a safe bet the answer here must be ‘not nearly so much’ given the societies in question are poorer, less densely populated (visible from settlement evidence) and less urbanized (same), but how much less so??Almost entirely speculative.
Building on Sand
So what can I tell you about the population of the ancient Mediterranean?
Not much!?In many regions, we can chart settlement data to get a sense of how the population changed, but not its absolute number.?Here, the basic is that we see almost everywhere slow growth, picking up pace in the Hellenistic period (except for some areas like Greece already heavily urbanized by that time), with a peak population reaches probably in the second century AD., and as we’ve discussed, a severe population collapse in the fourth and fifth centuries.
But turning that general sense into numbers – rate of change, absolute figures at particular dates??Outside of Roman Italy, the task is almost hopeless.?The best we can do are very rough order-of-magnitude figures, often with huge error-bars.?And those figures are in turn often a product of their own assumptions, creating a real danger for circular reasoning where, say, the population of Gaul needs to be low because we figure it was underdeveloped and because the population of Gaul is low, we figure it must have been underdeveloped.?Archaeology can backstop that doom-loop to a degree – pre-Roman Gaul was less urbanized than Roman Italy, for instance, in ways that are archaeologically visible – but here too we are refining a broad sense of the matter, not a specific estimate.
But all of this discussion brings me back around to a core observation here which is how little we know for certain.?Even in a case where you have responsible academic estimates across cultures, those estimates may not be comparable: if you wanted to compare the population of Roman Italy and late-classical or Hellenistic Greece, you could easy end up comparing 4.2m (De Ligt) with 3m (Corvisier and Suder) and say Italy has more or compare it with M.H. Hansen’s upper-bound at 6m and say it has less.?To be really crazy, you could compare Corvisier and Suder with Elio Lo Cascio’s figures for Italy and come to the opposite conclusion that Roman Italy had more than twice the population of Greece.17
Where that uncertainty ends up mattering most is in the efforts to do ‘macro-history’ that covers the ancient world by scholars that do not appear to have a firm grasp on the underlying weaknesses in what they perceive as ‘data.’?Economic historians in particular who assume that works like McEvedy and Jones,18 can be a useful tool to back-project ‘global GDP’ or estimate the productivity of ancient societies are fooling themselves.?Likewise, the notion that we can assess ‘elite overproduction’ in antiquity is a castle built on sand: we don’t even know to the nearest million how many people there were or to the nearest ten thousand how many ‘elites’ there were.?Significant changes in population are almost impossible to track in all but the most well-documented regions – a perhaps 25% increase in the population of Roman Italy really only became clearly archaeologically visible to us in the last 30 or so years and remember: Italy is the best archaeologically documented place on Earth.?In places that are not Roman Italy, it is not just possible but fundamentally certain that there could be double-digit percentage movements in population that are not now and may never become archaeologically visible to us.?We just don’t know.
What we do know is that basically all peoples at all times complain about useless, lazy aristocrats – the useless, lazy aristocrats complain about it most of all, so you will always find evidence of that complaining.?But if you project that complaining back into demographic studies built on these sorts of estimates, you are building on sand.
That’s not to say all of these studies are useless.?As a means of understanding the Roman economy, modeling based on demographic assumptions and what we know about Roman agriculture can be very helpful, so long as one keeps in mind how approximate it is.?Those are models build almost entirely on assumptions, not facts.?They are not data nor can they become data – they can never become stronger than the meager evidence that supports them.?Short of inventing a time machine, we are never going to have confident population estimates for most of the ancient world outside of strange places like Roman Italy (from 225BC to 14AD) and perhaps Roman Egypt.
But we have to approach our own meager powers and the meager powers of our evidence with due respect.?We can estimate, in a very rough order-of-magnitude sort of way, about how many people may have lived in antiquity, but for any approach that requires even extremely modest precision, the evidence simply will not support such efforts.
We have to be willing to admit what we do not and indeed cannot know.