CHAPTER 27
DEMOGRAPHY
.
The underlying demographic structure of the early Roman empire is only
dimly perceptible.1 By and large we lack not only reliable statistical evidence
for general demographic functions, but also the detailed local records that
prove invaluable for early modern Europe. What we do have, in abundance,
are impressionistic and often moralizing observations by literary sources;
but such remarks, as vague and inconsistent as similar statements by
modern lay persons, must always be considered suspect unless they can be
verified. Anecdotes are of similarly marginal demographic interest. For
these reasons we often have no choice but to fall back upon reasonable
conjecture: likelihoods, not truths.
Further, the Roman empire spanned a vast geographic range, and it
endured for centuries. To judge from early modern data, little uniformity
can be anticipated of it – considerably less, indeed, than the discussion in
this chapter may suggest.
Despite these handicaps, a picture is emerging. The Roman empire’s
demographic structure, to the extent we know it, broadly resembles most
populations before the modern demographic transition; in particular, it is
close to the norm for pre-modern Mediterranean societies, while displaying
no divergences that clearly anticipate the demographic transition. It goes
without saying that Rome’s demographic structure fundamentally conditioned the economic, social and political institutions of the Roman empire.
Roman demography can be approached in two ways. First, the population of the empire and of its regions can be examined for level, increase or
decrease, age and sex structure, and so on. Second, population can be
broken down into its three major demographic components: mortality, the
rate at which members passed out of the population through death; fertility, the rate at which new persons were born into the population; and migration, the rate at which persons entered or left the population through
physical relocation. These components jointly determined the general
structure and age distribution of the Roman population, as well as its
change over time.
1
On demography, see Newell ().
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
This chapter first examines the three components of Roman population, and then the empire’s gross population.
.
Almost all historians now assume that Roman life expectancy at birth was
approximately twenty-five years.2 This consensus rests less on ancient evidence, which is sparse and poor in quality, than on the reasonable conviction that, granted the general social and economic conditions prevailing in
the Roman empire, its life expectancy is likely to have lain near the lowest
levels attested for pre-modern populations. Of particular importance here
are data from India and rural China, which still in the early twentieth
century had life expectancies at birth in the lower twenties.3
. Probable life tables
No accurately measured population has ever had so low a life expectancy,
at any rate in normal times. Therefore, in order to understand what such
mortality implies, we must first turn to model life tables. These computergenerated models, developed to facilitate study of demographic history
and development, are based on historical data and describe ‘typical’ populations at various levels of mortality. Of the standard models, Model West
is the most generalized and widely applicable, and it is chiefly used below.4
Table gives mortality functions associated with Model West, level , in
which life expectancy at birth is years for females and . years for
males.5 In this life table, three columns of statistics are provided for
females and males aged exactly , , , and thereafter at five-year intervals
until age . The first column states the probability that a person of given
age will die before the next indicated birthday; thus, about sixty-two of one
thousand -year-old females will die before their twentieth birthday, and
about fifty-five of one thousand males. The second column shows the toll
such mortality would exact on representative ‘cohorts’ of one hundred
thousand newly born females and males; here the impact of high infant
mortality rates is especially clear. The third column gives average life
expectancy at successive ages; thus women aged have on average about
. years of life remaining, while their males counterparts have only .
years.
2
See, e.g., Parkin () –.
India: Coale and Demeny () – (data for ). China: Barclay et al. () – (data for
–).
4
On life tables and models: Newell () –, –. On the reasons for preferring Model West
for high mortality populations, see Coale and Demeny () , .
5
Coale and Demeny () . All data cited below on model life tables also derive from this volume
and are used by permission of Academic Press, Inc., and Prof. Coale.
3
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
Table : Model West, level : a life table for the Roman empire?
Females
Age
Males
Mortality
Cohort
Life exp.
Mortality
Cohort
Life exp.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Source: Coale and Demeny () . ‘Mortality’ is q(x), the likelihood that a person aged
exactly x will die before the next indicated birthday; ‘Cohort’ is l(x), the survivors to exact
age x; ‘Life Expectancy’ is e(x).
A caveat is required at this point. This model life table can give only an
approximate notion of normal Roman mortality experience. Because such
models are based on little empirical data for levels of mortality as high as
Rome’s, they are not entirely dependable, especially in two important
respects: the structural relationship between juvenile and adult mortality
levels, and the relative mortality levels of females and males.6 In any case,
mortality within the Roman empire must have fluctuated considerably
from period to period, region to region, and probably also from class to
class. A range of ten years in life expectancy would not be unusual; thus,
normal Roman life expectancy at birth is perhaps more satisfactorily set in
a broad range from twenty to thirty years.
6
On infant mortality, see Bagnall and Frier () – n. , with further references. On sex
differentials, see below.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
The best surviving ancient evidence, though it is often difficult to interpret, generally supports modern assumptions about Roman mortality. A
good example of the problems with Roman sources is a schedule for calculating the tax value of annuities, a document commonly called Ulpian’s
life table but in fact apparently created by imperial bureaucrats; it is preserved in a juristic text dating from the early third century .. Ulpian’s life
table appears to give figures for adult life expectancy, and these figures are
broadly consistent with Model West, level , in which female life expectancy at birth is . years, and male about ..7 Ulpian’s life table thus
implies mortality rates even higher than those usually presumed for the
Roman empire; but the statistical peculiarities of Ulpian’s schedule are so
obvious that a degree of continuing scholarly caution may well be justified.8
On the other hand, much the same result emerges from what is generally conceded to be by far the best surviving demographic source for ordinary subjects of the Roman empire: the three hundred census returns,
containing entries for more than eleven hundred persons, filed in Roman
Egypt during the first to third centuries .. Despite uncertainties, the age
distribution in the census returns strongly suggests that overall Egyptian
life expectancy at birth was in the lower twenties, probably between and
years.9 There is no reason to suppose that this result is entirely coincidental.
Skeletal evidence from Roman cemeteries has rarely been subjected to
accurate demographic analysis, and in any case the obstacles to determining the age and sex of skeletons remain formidable. Two well-studied
fourth-century cemeteries from Pannonia do both yield mortality data that
closely support Ulpian’s schedule.10 However, although close study of
Roman cemeteries may eventually greatly augment our knowledge of
demography, the results thus far are of limited value except as to health and
the causes of death. Of exceptional importance, in this regard, are the
skeletons buried at Herculaneum when Mount Vesuvius erupted in .. ;
although at best they represent only a cross-section of a well-off Italian
population, they are already proving useful to demographic anthropologists.11
One final source is far more controversial. Especially in the Latin-speaking West, Romans often include on epitaphs the decedent’s age at death;
this practice was almost universal in North Africa, but markedly less
common in Europe. Although the corpus of surviving inscriptions is
7
Aemilius Macer ( ad Leg. Vic. Hered.), Dig. .. pref. (quoting Ulpian), with Frier ().
Saller () -, summarizing recent scholarship.
9
Bagnall and Frier () –, summarizing discussion at pp. –; a low intrinsic growth rate
is assumed.
10
Acsàdi and Nemeskeri () –; Frier (b). Compare Parkin () –, on palaeo11
demography; with Jackson () –, on palaeopathology.
Bisel ().
8
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
numerically impressive (some , from Europe, an additional ,
from North Africa west of Egypt), gravestones are today widely and rightly
treated with suspicion.12 In the case of the European inscriptions, no life
table based on all or part of them is even remotely plausible. The inscriptions invariably underrecord juvenile mortality, but that could be corrected.
More seriously, they grossly exaggerate mortality among young adults, evidently because the decedent’s tender years often influenced the decision to
include age at death; further, the mortality rates for older adults are patently
skewed by age exaggeration. Thus every portion of these epigraphic life
tables is incurably biased.13
Roman North Africa is altogether different. Unlike in Europe, African
epitaphs almost invariably give decedent’s age at death, so problems of
selectivity in including age at death are less severe. Although the African
epitaphs underrepresent juveniles, they produce credible mortality rates for
males aged to , and for females aged to ; and, as table shows,
these mortality levels are reasonably comparable to Model South, level ,
in which both sexes have a life expectancy at birth of about . years.14
Male mortality virtually duplicates the model; female mortality is generally
similar to the model, but significantly higher during the peak years of childbearing, ages to – perhaps a sign of high fertility. In later ages,
however, mortality rates are artificially lowered owing to age exaggeration,
a phenomenon that apparently begins somewhat later for women than for
men. Since tombstones were moderately expensive, these epitaphs indicate
that high levels of mortality obtained also among the urban well-to-do.
Empirical evidence thus generally supports the modern consensus that
average life expectancy of Romans at birth was normally about years, or
perhaps even slightly lower.
. Causes of mortality
The concepts of ‘stable’ and ‘stationary’ population are powerful theoretical tools in demography. If a population’s birth and death rates remain
unchanged over many generations, and if migration is negligible, that population becomes ‘stable’, with a constant age-structure and rate of growth.
If birth and death rates are also identical, then this stable population is also
‘stationary’; its numbers remain unchanging. Though in fact no historical
society has ever been exactly stable (much less stationary), pre-modern
12
See just Hopkins () and Parkin () –. Szilagyi compiled this epigraphic evidence in
AArchHung for to and to .
13
Suder () unsuccessfully attempts to evade these problems.
14
Table does not use the inscriptions from Carthage and Mauretania, which diverge from the
pattern found elsewhere in Africa. Model South presupposes relatively low mortality from ages to
; see Coale and Demeny () .
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Table Reported mortality rates in Roman North Africa
Females
Age
Males
Inscriptions
Model South,
level
Inscriptions
Model South,
level
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Source: African mortality rates – q(x) – calculated from statistics in Szilagyi (–), with
five-year moving averages used for inscriptions giving age at death. Life table: Coale and
Demeny () .
populations can often be approximately described through these concepts.15
Under the conditions of Model West, level , a stable population, if it is
stationary or slowly growing, will have annual death rates of about forty
per thousand for women, forty-four per thousand for men; more than half
of all deaths will be of children under ten. These death rates are very high;
in early modern Europe the death rates were rarely higher than thirty-five
per thousand.16
Death in the Roman world followed a seasonal pattern also found in the
early modern Mediterranean: highest in late summer and early autumn,
when infectious diseases took their heaviest toll; lowest in the cooler winter
months.17 The major natural causes of death probably did not differ much
from those prevailing in early modern Europe. Although many diseases
have a history of their own – now virulent, now abating – the great
15
16
For details, see Bagnall and Frier () –.
Wrigley and Schofield () –.
Norberg () –, cf. Lassère () –; in modern populations deaths are usually highest
in mid-winter. The seasonal pattern of Roman births almost exactly reverses that for deaths: Norberg
() –.
17
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
scourges are certain to have been, first, the numerous ‘fevers’, including
typhus, typhoid fever, Malta fever and malaria; second, pulmonary illnesses,
especially the forms of pneumonia and tuberculosis. In normal circumstances, these causes were probably immediately responsible for around
per cent of all deaths. Also undoubtedly significant were dysentery and
diarrhoea (especially for infants); cholera; gangrene; scurvy (especially in
times of want); and, less frequently, rabies, tetanus and anthrax. Ancient
medical writings describe most of these illnesses, though for advanced
cases doctors could offer little more than comfort. The evidence for
bubonic plague, measles, smallpox (before .. ), influenza and syphilis
is less certain.18
Casual violence or accident, including tainted food and drink, also
carried off a substantial number of Romans. A man of from Ephesus
died of a haemorrhage after drinking a massive dose of wine; an African
widow records that her -year-old husband had been ‘deceived by a bull’.19
Military deaths, by contrast, were of little statistical significance in peacetime; the Roman army constituted less than per cent of the empire’s population.
Why did Roman death rates remain so high? The ineffectiveness of
medical science explains little; medicine had no measurable impact on
death rates until at least the early eighteenth century. Four interconnected
reasons may perhaps be given for high Roman mortality.20 First, poor nutrition, conditioned primarily by the low level of real wages, rendered most
Romans susceptible to illness; and although the Herculaneum skeletons
show that at least the affluent were well nourished, they also provide some
alarming evidence of lead poisoning. Second, sanitary standards were poor
especially as to the disposal of human waste and garbage; large cities in particular remained fetid despite the spectacular feats of Roman engineers in
providing fresh drinking water. (Rome alone probably produced about
million cubic metres of human waste each year, a fact worth remembering
when we read of Romans bathing in the Tiber. Indeed, the medical writer
Galen specifically warns against eating fish from the Tiber.) Third, Roman
urbanism implied large and compact settlements linked by swift communications, and thus provided a ready network for infectious diseases to take
hold and spread. Fourth, unlike early modern nation-states, the underbureaucratized Roman Empire could not or would not take the draconian
measures required to quarantine and eradicate pestilence.
The Roman empire’s vast expanse helped insulate its total population
18
See generally Grmek ().
Ephesus: Meillier (). Africa: CIL . (‘a tauru deceptus’). On banditry, see Shaw
(a).
20
These reasons are ‘extrinsic’ to the demographic structure. In section below, it is suggested that
high fertility (an ‘intrinsic’ reason) was ultimately responsible for high Roman mortality.
19
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
against ‘great crises of mortality’ caused by plague or famine; what for an
early modern state would have been a national catastrophe, for the Romans
was essentially a regional event, though no less severe in its consequences
for local populations. The balance of food production was always precarious, and the even distribution of bulky staples like cereals was hampered
by the high cost of overland transport; even in Egypt, usually a large net
producer of cereals, food shortages occurred sporadically, and in most
other areas of the empire they were fairly frequent.21 At Rome, pestilence
erupted so frequently that literary sources pay it slight heed. Even natural
catastrophes usually provoked new outbreaks; for instance, the eruption of
Mount Vesuvius in was soon followed by pestilence (perhaps cholera)
in Rome.22
However, apart from vague references to widespread epidemics under
Domitian and Hadrian,23 the early empire as a whole was spared a ‘pandemic’ until .. , when the army of Avidius Cassius, returning from
Parthia, brought with it what was probably smallpox, now for the first time
establishing a permanent foothold in the Mediterranean basin. This plague,
the first of many that enervated the later empire, raged for a quarter
century; in , at the height of its second outbreak in Rome, an eyewitness says it caused two thousand deaths per day.24 The Malthusian checks
had begun.
. The age structure and differential mortality
Even in more normal times, high mortality rates produced a youngish population. In the stationary population of Model West, level , the average age
is . for females and . for males; in the Egyptian census returns the
average age is . for females and . for males, a fairly close match.
Indeed, as table shows, the general age distribution of the Egyptian population was quite close to the model: about per cent of the population
was less than , some per cent was to , and only slightly over per
cent was or older. The same age distribution seems to be emerging
among the Herculaneum skeletons.
There is little firm evidence for major class-related differences in Roman
mortality rates; studies of ‘differential mortality’ on the basis of epitaphs
are highly problematic.25 If experience in the Americas is any guide, Roman
slaves probably had lower than normal life expectancy, and the
Herculaneum skeletons may confirm this; but slaves represented considerable capital investment, and hence were treated with some care even in the
21
22
23
Garnsey, Famine.
Suet. Tit. .; cf. Dio, . ..
Dio . ; HA Hadr. ..
Dio . .–. Smallpox: Littman ().
25
E.g. Suder () –; cf. Salmon () –. See also section below on the high mortality in larger Roman cities.
24
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
Table Age distribution of Egyptian population
Females (per cent)
Age
–
–
–
–
–
–
–
–
–
Number:
Males (per cent)
Census
returns
Model West,
level
Census
returns
Model West,
level
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Source: Census returns: Bagnall and Frier () (figures updated). Life table: Coale and
Demeny () , .
Republic. Slaves may well have been better off in this respect than subsistence peasants of free status.26
At the opposite end of the social scale, it has been observed that the
Roman Senate, which in the empire normally had about six hundred
members, was maintained through annual entry of twenty quaestors each
aged about .27 This implies a life expectancy of about thirty years for senators aged , which is consistent with a male life expectancy at birth of
about thirty years (Model West, level ). Senators may thus have been
somewhat more advantaged than the general population. Ironically, high
status among the Romans usually meant urban residence during most of
the year, and hence increased exposure to urban diseases that did not
respect position. The incidence of tuberculosis among Pliny the Younger’s
friends is eloquent in this regard, as is Tacitus’ report that numerous senators and equestrians died in the pestilence at Rome following the Great Fire
of .28
It is also hard to make out a sex-based differential in mortality. However,
comparative evidence indicates that, in the least developed pre-modern
populations, males may frequently have enjoyed somewhat longer life
expectancy than females, a situation that the demographic transition has
since reversed. The situation in Roman Egypt, the best-documented
26
Evans ().
Hopkins, Death and Renewal –; cf. Duncan-Jones, Structure – (on the album of Canusium).
28
Pliny, Ep. .., ..; Tac. Ann. ..–. On tuberculosis, cf. CIG (Smyrna), IG 2
(Epidauros).
27
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
ancient population, is unclear on present evidence, but the census returns
at least suggest that males had a similar or slightly higher life expectancy
than females, so that the sexual imbalance normally prevailing at birth
(about male newborns for every females) continued into adult
life.29 If this inference is correct, then Model West, level , should be used
for males, rather than level .
. The perception of aging
Under the conditions of Model West, level , only one newborn in eight
survives to age . A Roman who reached this age was already ‘old’, senex;
but the word senex was also applied to much younger people. Cicero, for
instance, once describes a man of as senex.30
The Egyptian census returns show that at least some ordinary Roman
subjects were able to determine their ages accurately throughout their
lives.31 The epitaphs imply, however, a more complex situation, further
complicated by the fact that the ages come not from decedents but from
their survivors. African epitaphs may indicate that accurate age remained
part of a person’s ‘identity’ until about age . However, both in Africa and
elsewhere, after age the reports of age at death are increasingly given in
multiples of five.32 In Africa this pattern of age-rounding obtains for nearly
two-thirds of all epitaphs; in Rome and the European provinces the phenomenon is only slightly less pronounced. Preference for certain digits followed regional patterns throughout the empire; thus, in Africa there was
also a proclivity toward ages ending in one, while two was preferred in
Rome, and the Egyptians (on mummy labels) liked six. Such digit preference argues that those who erected the epitaphs either were not concerned
to give exact ages, or were unsure of the decedent’s exact age, perhaps
because the decedent had also been unsure. The second explanation is by
no means impossible. In one especially notorious case, a well-off Egyptian
landowner is repeatedly given inconsistent ages in a series of dated documents.33
For the African epitaphs, comparison with the model life tables suggests
that after age the Africans, both male and female, lived about nineteen
‘psychological years’ for every ten calendar years. Doubtless exaggeration
was assisted by age-rounding, which allowed age to skip ahead by pentads.
In Africa nearly per cent of all adults are reported to have died at age
and over, and nearly per cent are centenarians; but for a stationary population the comparable model life table gives and per cent, respectively.
29
30
31
33
Bagnall and Frier () –.
Censorinus, D. N. ; Cic. De Or. . (of his teacher L. Licinius Crassus). Cf. Suder ().
32
Bagnall and Frier () –.
Duncan-Jones, Structure –.
Boak and Youtie () (Aurelius Isidorus, fourth century ..).
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
Roman Africa took pride in the number of its elderly,34 and age exaggeration was doubtless more pronounced for this reason. But similar, though
less extreme, exaggeration occurred elsewhere in the empire; in .. a
census of Italy’s eighth region produced a bevy of centenarians, none of
whose ages are credible.35 The old lived on, venerated sincerely, if somewhat artificially, in a society where the old were rare.
.
Harsh mortality rates placed a considerable burden on Roman society, and
above all on the reproductive capabilities of Roman women. Literary
sources, reflecting the perceptions of the status élite, often give the impression that Roman families commonly had few or no children.36 But this
cannot have been generally true, since even a modest fall-off in the required
birth rate would soon have caused a precipitous decline in population.
. The gross reproduction rate
If a population is to endure over a long period, the minimum requirement
is that each generation of women replace itself. In demographic terms, the
net reproduction rate (NRR) must be ., meaning that on average each
woman reaching menarche bears one daughter who also reaches menarche.
More useful, however, is the gross reproduction rate (GRR), a more
abstract concept. It is calculated by aggregating the number of female
births per living woman at various ages throughout the period of female
fertility (by convention, ages to ); it thus gives the number of daughters a woman will bear if she survives to age and bears daughters at an
average rate for women her age. When the effects of mortality on adult
women are then weighed in, the GRR is converted to the NRR.37
Under the mortality conditions of Model West, level , a GRR of about
. is needed for a stationary population, and a GRR of about . for a
stable population growing at an annual rate of . per cent.38 This means
that, if the Roman population was stationary or moderately growing, the
average woman who reached menopause probably bore at least five to six
children altogether. These rates are inexorable. For example, if the GRR
had been not . daughters but only ., then the Roman population, when
34
Sall. Cat. ., with Lassère () –.
Pliny, HN .–; Phlegon of Tralles, FGrH F ; cf. Pliny, Ep. ... on Tifernum
Tiberinum. Cf. Parkin () –.
36
See, e.g., Eyben (–), esp. , with sources.
37
On measuring fertility, see Newell () –. Fertility rates count only live births, not miscarriages or stillbirths.
38
Coale and Demeny () , assuming a mean age at maternity of .. This age is about right
for Roman Egypt: Bagnall and Frier () –.
35
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
it stabilized, would have halved every century. But no modern historian
contemplates a decline of such magnitude, and most suppose that the
empire’s population was stationary or slowly growing. It must follow that
the actual long-term GRR for the empire as a whole (though, of course,
not necessarily for every class and region within it) remained close to, or
exceeded, the GRR required for a stationary population. If literary sources
appear to paint a different picture, they are not to be trusted.
The projected GRR for the Roman empire may seem high, but it is in no
respect extraordinary for a pre-modern population. For example, mideighteenth-century France, under mortality conditions closely similar to
those posited for Rome, still had a GRR above . despite a late age for
women at marriage; and the least-developed contemporary countries frequently have even higher fertility rates.39
The burden of this fertility was not, however, distributed equally among
adult Roman women. Although non-marital fertility (including births to
slaves) was not insignificant, free married women undoubtedly had considerably higher fertility rates than unmarried women, evidently because of
the strong and enduring cultural link between marriage and procreation; as
the doctor Soranus candidly observes, ‘women usually are married for the
sake of children and succession, and not for mere enjoyment’.40 In Roman
Egypt, for example, some per cent of all births were within wedlock, but
only around per cent of all Egyptian women aged to were married
at any one time; this implies marital fertility rates about four times higher
than non-marital rates.41
The pattern of Roman fertility was chiefly determined by two factors:
marriage customs, especially the age of free women at marriage and the
probability that they remained married until menopause; and the methods
by which fertility was controlled within the ancient world.
. Marriage
Roman women married early, and thus were able to bear legitimate children
during all or almost all of their peak reproductive period. Close studies of
epitaphs from the western empire indicate that women generally married
in their late teens, although the best evidence (the shift in commemorators
from parents to husbands) is necessarily somewhat indirect.42 Much the
same pattern is also found in the Egyptian census returns, where it can be
more exactly studied. Women begin to marry at age or , shortly following menarche; per cent or more have married by age , and by
39
For sources, see Bagnall and Frier () –.
Soranus, Gyn. .. See, e.g., Treggiari (a) –; Dixon () –.
41
Bagnall and Frier () –, –. Non-marital fertility is concentrated mainly among slave
42
women.
Shaw (b); Saller () –.
40
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
nearly all women are married or previously married.43 Maternity began
soon after marriage; in Egypt, almost a quarter of all legitimate children
were borne to mothers under age .
This custom of early female marriage closely resembles that prevailing
in pre-modern Mediterranean populations, and it is concordant with a
marital regime in which at least the first marriage of women is usually controlled by their families. Among the Roman élite, marriageable daughters
were often treated as dynastic pawns, and their first marriages were even
more accelerated.44
Both inscriptions and the census returns show that men normally
married somewhat later than women, usually in their mid-twenties, so that
husbands were not infrequently some seven to ten years older than their
wives.45 However, at least in Egypt, the male marriage pattern is more
complex than the female: men marry over a longer period, and though
many begin to marry in their late teens, others may remain unmarried until
as late as their forties.
Lifelong celibacy was rare for freeborn women; Rome had no spinster
class. A survey of African epitaphs for women of marriageable age showed
that, in those cases where it was possible to determine marital status, nearly
per cent were or had been married; and of those who were evidently
unmarried, half were still under age .46
For female slaves and freedwomen, the problem of marriage was more
complex; but many entered into informal unions (concubinage or contubernium), either with those of similar status or, less often, with their masters
or patrons. Soldiers and sailors, forbidden to marry during their protracted
term of service, also often resorted to concubinage.47 The frequency of
concubinage meant that illegitimate children were common: in Egypt,
perhaps around to per cent of free births, although higher locally and
especially in villages. For both mother and child little social stigma attached
to illegitimate birth.48
The Romans themselves were relentlessly monogamous, rarely even
combining concubinage with marriage, and they also enforced reasonably
strict rules against incest and incestuous marriage. Thus, in Egypt
brother–sister marriage remained common (in the census returns, about
one-sixth of all marriages) until the Egyptians were made Roman citizens
in .. /.49
High mortality rates meant that many marriages were broken by the
43
Bagnall and Frier () –. On menarche and menopause in the ancient world, see
44
Amundsen and Diers (), ().
Syme (); Treggiari (a) –.
45
46
Saller (a); Bagnall and Frier () –.
Lassère () –.
47
48
Treggiari (a); Campbell ().
Egypt: Youtie (). In general: Rawson ().
49
Bagnall and Frier () –; see generally Treggiari (a) –. In the West, close-kin
marriages (e.g. with first cousins) were rare even though permitted: Saller and Shaw (a).
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
death of one spouse; thus if a man aged married a woman aged , the
figures in table imply better than one chance in four that one or both
spouses would die within ten years. Divorce, in principle unrestricted by
law and available on demand to either spouse, was evidently frequent
among the upper classes, but its incidence among ordinary Romans is
debated. However, in the Egyptian census returns divorces are not unusual,
suggesting little social restraint on divorce among the lower classes as
well.50
When marriages were broken by death or divorce, the former spouses
not infrequently remarried. Among the upper classes the Augustan marriage legislation exerted pressure to remarry, and literary sources indicate
that remarriage was normal, particularly for women still of childbearing
age. Among ordinary Egyptians remarriage is also quite common, particularly after divorce; but women apparently seldom remarry after age ,
while men remarry, to increasingly younger wives, throughout their lives,
and by age all surviving males appear to be either married or previously
married.51
. The fertility pattern
Before the modern demographic transition, all accurately measured populations have a characteristic pattern of marital fertility rates across the years
from menarche to menopause. This pattern is called ‘natural fertility’
because its age distribution – its shape – is determined almost entirely by
the ordinary level of adult female fecundity (potential fertility): marital fertility rates of women decline gradually in their s and s, and then sharply
in the s, as a direct function of declining fecundity as women age.
Therefore the shape of marital fertility is evidently not influenced by the
attempts of couples to limit family size after reaching what they consider to
be a sufficient number of children. By contrast, today, in the aftermath of
the fertility transition, family limitation is obviously a pervasive aspect of
marriage; childbearing is usually concentrated in the early years of marriage,
and marital fertility rates decline much more rapidly than does fecundity as
women age.52 Although this may seem a mundane matter, recovering the
age distribution of marital fertility is crucial to a deeper understanding of
Roman demography.
The only surviving evidence as to the Roman pattern of marital fertility
comes from the Egyptian census returns, which apparently reflect a fairly
stable population with a small annual growth rate. Table gives both overall
50
Treggiari (a) –; Bagnall and Frier () –.
Upper classes: Humbert () –. Egypt: Bagnall and Frier () –.
52
See Frier (), with bibliography, esp. Coale (), on both of which the discussion below
draws heavily.
51
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
Table Reconstructed fertility rates, Roman Egypt
Reconstructed
fertility rates
Age
–
–
–
–
–
–
–
–
Standardized fertility
rates (– = )
All women
Married women
Egypt (married)
Natural fertility
.
.
.
.
.
.
.
.
.
.
.
.
Source: Fertility rates for all Egyptian women: Bagnall and Frier () ; rates for all
married Egyptian women and natural fertility rates: Frier () . Fertility rates are
expressed as annual births per one thousand women of the indicated age.
and marital fertility rates derived from the reports in the census about the
difference in age between a mother and her children.53 For each successive
pentad of a woman’s life, the corresponding fertility rates express the
annual number of births per one thousand women. The mean age of
maternity in the census returns is about years, typical of populations
with natural fertility; the mean age of paternity is much higher, probably
about years, reproducing the wide age gap between husbands and wives.
Graph represents the values from table after standardization: the fertility rate for married women aged to is assigned a value of , and
fertility in subsequent pentads is measured as a percentage of that for ages
to . As this figure shows, marital fertility corresponds closely to the
normal pattern for natural fertility. Therefore ordinary Egyptian couples,
in so far as we can know of their behaviour, did not anticipate the modern
fertility transition; they did not attempt to limit family size, and so women
continue bearing children at a relatively high but declining rate well into
their forties. Although this pattern is confirmed only for Roman Egypt,
there is no appreciable likelihood that it, or a pattern closely similar to it,
did not prevail generally in the Roman empire.
The marital fertility rates in table may seem surprisingly low. Even in
their peak period of fertility (ages to ), married women bear children
in only one year out of every three, comparable to most Mediterranean
populations before the fertility transition, but far below the highest rates
that are known to be socially sustainable. The Egyptians could accept such
53
Overall fertility: Bagnall and Frier () –. Marital fertility: Frier () –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Graph Egyptian fertility rates and natural fertility pattern
low rates because, like almost all pre-modern populations, they distributed
the heavy burden of childbirth as widely as possible among fertile women:
not only did women marry early, but all, or virtually all, women married.
But such low fertility rates force us to confront what is perhaps the central
enigma of Roman demography: how did the Romans prevent a population
explosion?
. Fertility control
Although the shape of the natural fertility pattern is nearly constant in premodern populations, the actual level of fertility varies extremely widely;
some populations (such as the United States in the early nineteenth
century) have elevated fertility levels, while others (such as rural China in
the early twentieth century) have very low levels. The cause of this variation is, in the main, the relative effectiveness of controls on fertility; but,
particularly within marriage, these controls do not have a modern form.54
If a population with the Roman empire’s general structure of mortality
and nuptiality succeeds in maintaining a stationary or slowly growing population, the reason will lie about as much in voluntary or social
restraints on maximum fertility, as in the harshness of its death rates.55
Ancient sources leave no doubt that Romans, or at least some Romans,
were genuinely interested in controlling fertility through contraception and
54
See, e.g., Wood (), with a survey of modern scholarship.
55
Weiss () –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
abortion. Medical writings in particular contain numerous discussions of
both practices (which were often freely confounded); and neither was
visited with substantial moral disfavour or legal sanction during the early
empire.56 To be sure, the efficacy of the suggested methods is often open
to question, although one recent study has argued strongly in their favour.57
In general, suggested abortifacients were probably more effective than
contraceptives, though many such methods risked the mother’s life if not
dosed exactly.
But even if Romans had access to dependable methods of contraception and abortion, the more significant demographic question is how, and
by whom, these methods were used.58 Here comparative evidence is decisive: within marriage, contraception and abortion, as ‘direct’ methods of
fertility control, are emphatically associated with conscious family limitation and hence with the modern fertility transition. No general population
is known to have practised family limitation before the fertility transition,
and, as we have seen, Rome was apparently no exception. If Romans made
use of contraception and abortion, they are likely to have done so primarily in the context of non-marital fertility, where the strong cultural link
between marriage and procreation did not obtain.59 Non-marital fertility
includes all illegitimate births to free and slave women, irrespective of
whether such births result from non-marital relationships (such as concubinage or contubernium) that could have long duration.
If not by contraception and abortion, then how? In pre-modern populations, marital fertility was apparently controlled, for the most part, by
‘indirect’ methods, such as breastfeeding, that act to delay post-partum
pregnancy across the entire span of female fertility. These indirect methods
do not limit family size, but instead ‘dampen’ fecundity in a fairly even way.
Prolonged breastfeeding was clearly widespread in the Roman world; the
doctor Soranus, for instance, recommends weaning after eighteen to
twenty-four months. Only wealthy families made use of wet-nurses.60 In
Egypt, wet-nursing contracts, which doubtless reflect normal breastfeeding practices, usually last for two years. Further, many of these contracts
also enjoin the nurse from sexual intercourse, indicating that abstinence
during lactation was considered desirable or even mandatory – a folk taboo
that Roman doctors also commend and that is common in many traditional
societies.61
In addition to direct and indirect forms of fertility control, the Romans
also resorted to infanticide or exposure.62 Exposure differs from infanticide
56
The modern scholarship is extensive; see Parkin () –, summarizing the debate.
58
Riddle ().
Frier () – (criticizing Riddle).
59
This is, at least, the likeliest explanation; see Alter ().
60
Soranus, Gyn. .; compare Parkin () –. On wet-nursing: Bradley () –.
61
See Bagnall and Frier () –, with further references. Post-partum abstinence is recommended by, e.g., Soranus (Gyn. .) and Galen (De San. Tuenda . .–, = CMG ., p. ).
62
See esp. Harris (), with a survey of the massive bibliography.
57
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
in that the exposed child is given at least a theoretical chance of being taken
in by a stranger – normally to be raised then as a slave. Infanticide and exposure were not commonly considered immoral, nor were they made illegal
until the later empire; even the upper classes often put newborn children
to death if they were deformed or unlikely to survive, and the lower classes
allegedly did so also for economic motives or on the basis of sexual preference. The frequency of the practice cannot be determined, but it was
plainly not negligible. Literary sources suggest that female newborns were
more likely to be exposed than males; but surviving data on the sex ratio
are inadequate to demonstrate any large effect of the practice.63
In any case, the Egyptian census data indicate that exposure was not
used to limit families; it was perhaps more commonly associated with nonmarital fertility, including adulterous births, in cases where contraception
or abortion were unavailing. Even within marriage, such direct and conscious forms of fertility control had their purposes, though not to limit
family size but rather to ensure a safe interval between successive births;
the dominant aim was to protect the health of the mother and of her earlier
children, which were placed at risk by too close a spacing. The Romans
were clearly aware of the risk.64
Large families were certainly not rare in the Roman world. The Italian
woman honoured at consecutive secular games for bearing ten children was
perhaps singular less for this fact than that all of them were still living; thus,
an African widower records quite in passing that his wife bore him twelve
children.65 But the absence of family limitation becomes apparent in the
Egyptian census returns, where we can examine its consequence more concretely. The most striking feature, largely produced by the randomizing
factor of heavy infant mortality, is the enormous variance in the number of
surviving children. Couples with as many as eight children are attested,66
but large numbers of surviving children are infrequent. In complete or
nearly complete returns, nine-tenths of married couples declare three or
fewer surviving children, and the average number of children declared is
less than two. Ancient sources that express preference for only two children presumably discount already for infant mortality.
. The level of fertility
In the end, although much can be surmised about Roman fertility, data are
lacking to determine its precise level. Fertility rates in the general popula63
Parkin () –. On Egypt, Bagnall and Frier () –, –.
See Gourevitch () .
65
Mart. .; CIL (Hippo Regius). Other large families: Treggiari () –.
66
BGU i = WChr , where Eirene gave birth at ages , , , , and (the age of two
other children is unknown); and doubtless there were other children, now deceased or departed.
64
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
tion were undoubtedly very high by modern standards; for example, in a
well-studied group of female skeletons from a British cemetery, adult
women had borne an average of about . children before their deaths, well
above what is required to maintain a stationary population under conditions of high mortality (about .).67 But such samples are usually too
small to permit generalization.
But mortality levels were also so high, the Roman population was fairly
delicately poised between the risks of under- and overpopulation; in this
respect, of course, it resembles all pre-modern societies. What is astonishing about these societies is that, irrespective of their relative mortality
levels, most had long-term intrinsic growth rates that lay between . and
. per cent per year; that is, their birth and death rates were usually closely
balanced, and population therefore grew slowly.68 The age distributions in
the Egyptian census returns are most consistent with an annual growth rate
around . per cent, sufficient to double the population every three and
one-half centuries.69 Such a growth rate would not have been difficult to
attain even under Roman mortality conditions, and there is no reason to
believe that the early Roman empire fell short of it.
However, the issue here is more complex than it may seem. In general,
moderate population increase was probably conducive to growth in traditional economies, and hence, of itself, desirable. At some point, however,
social conventions that mandated early and universal marriage could initiate a cruel cycle, first of population growth that outstripped resources, and
then of an offsetting increase in mortality rates, until an equilibrium was
eventually restored. France of the ancien regime may have approximated a
society of this type: its population already probably far above optimum, its
birth and death rates in any case much higher than in its northern European
neighbours.70 Such a population had apparently fallen into the grip of
Malthusian constraints, although exactly how this could have come to pass
remains among the deep mysteries of historical demography. What should
be stressed, in any event, is that this may well have been the normal fate of
pre-modern populations, a fate from which only few escaped.
Taken at its strongest, our evidence implies that the Roman empire was
not among these lucky few. Roman mortality and fertility rates, to the
extent that we can reconstruct them, are comparable to those of most traditional societies, at the high rather than the low end of the usual premodern range. Further, the Roman modes of fertility control, if viewed as
67
Wells () – (Cirencester, fourth century ..; thirteen women). In the model, average
number of births per woman who reaches age is calculated from table by this formula: . times
68
,, divided by female cohort survivors to age .
Livi-Bacci () –.
69
Bagnall and Frier () –.
70
Wrigley and Schofield () xxiv–xxv, –, ; and see generally – on the demographic
transition. (France had a late age of marriage, but very high fertility rates after marriage.) Rural China
in the s may parallel Rome even more closely: Barclay et al. ().
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Malthusian preventive checks on population growth, were probably less
responsive to short- and perhaps even long-term demographic changes
than those used in early modern Europe. They may well have been effective
in maintaining population equilibrium only after mortality had reached relatively elevated levels.
For Rome, therefore, the fundamental demographic risk was probably
not depopulation, but overpopulation, at any rate so long as its economy
was unable to generate new resources to sustain a growing population. But
literary sources of the early empire give exactly the opposite impression,
that population decline was the imminent danger; and such sources seem
directly reinforced by the Augustan marriage laws, which employed a
variety of sanctions and rewards to encourage marriage and childbirth.71
The contradiction is probably just a function of social perspective, since
literary sources are chiefly concerned with reproduction among the upper
classes, who were also the main target of Augustus’ legislation. In this tiny
élite, the failure to reproduce may indeed have been a pervasive difficulty;
it has been observed, for instance, that three-quarters of senatorial families disappear entirely after just one generation.72 Granted relatively more
favourable mortality rates in the élite, their failure to reproduce themselves
is indeed problematic. The likeliest explanation is that, among the privileged, more individualistic conceptions of marriage had developed, leading
to the widespread use of direct methods of fertility control (contraception
and abortion) in an effort to limit family size. There are clear historical parallels for this: long before the onset of the general fertility transition in
Europe, some small but affluent social groups were already consciously
limiting their families.73
Upper-class attempts to ‘limit the number of their children’ (finire
numerum liberorum) occasioned intense resentment. The social situation is
perhaps best captured by the satirist Juvenal, who bitterly observes that
poor women must inevitably give birth, while wealthy women resort to
drugs inducing sterility or abortion.74 Nonetheless, the individualistic antinatalism of the privileged probably did not extend very far down the social
scale; the municipal aristocracies of Italy, for instance, display remarkable
continuity during the early empire.75
In the late first century .., under imperial patronage, alimentary
schemes were set up providing financial subvention to parents who raised
their children. These schemes were soon copied by wealthy private citizens.
But the scale of such programmes was too small to have effected much,
and they were mainly confined to Italy.76
71
72
73
75
Treggiari () – (Augustan laws), – (failure to reproduce).
Garnsey and Saller, Empire () , –; compare Hopkins, Death and Renewal –.
74
Livi-Bacci ().
Tac. Germ. .; Juv. Sat. .–.
76
MacMullen, Social Relations , n. .
Duncan-Jones, Economy –, –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
. Mortality, fertility and household structure
Mortality and fertility rates operate as generalized probabilities, not as predictions for individual cases, which inevitably can vary quite widely from
any norm. The large range in numbers of surviving children, discussed
above, is a salutary example of this point: the toll of infant mortality fell
with grievous force on some families, while leaving others unscathed. But
the point has wider implications.
The Egyptian census returns provide the only secure evidence as to the
ordinary household size and structure in the early empire.77 In complete or
nearly complete returns, the average household was smaller than might
have been anticipated: about . family members. But again there is
extraordinary variation. Some per cent of attested households are of a
‘simple’ form: persons living alone, co-resident siblings, or conjugal
(nuclear) families. The remaining households are of ‘complex’ form: either
extended through the presence of near relatives, or with multiple coresident families. Such complex households could become extraordinarily
large; the largest have upwards of twenty family members, and sometimes
slaves or lodgers as well. It is likely that, in Egypt as a whole, at least threefifths of principal family members lived in complex households. This, too,
was characteristic of Mediterranean populations in the early modern
period.
Particularly in villages, the strong preference for complex households
clearly reflected cultural preferences as much as economic necessity. In the
census of /, one village household, by no means unusual, registered
four brothers, their wives and children, a total of nineteen persons; but a
separate return shows that in the same village the four brothers owned considerable other property, including a house, left standing vacant.78
The fortuitous impact of demographic probabilities meant that households swelled and contracted erratically from generation to generation.
Conjugal families, for instance, almost always are not young couples with
their children; rather, they result from attrition, through the death of
parents and other near kin, and so the couple are usually at least middleaged. In a few cases where we have successive census returns from the same
household, it is clear that a family could experience difficulty reproducing
itself into the next generation.79 On the other hand, because parents, too,
might fall victim to high mortality, orphans were not rare; and it is actually
common for households to contain children, even adult offspring, from
former marriages. In one remarkable household, a family of renters from
77
The data in this section derive from Bagnall and Frier () –; see also Saller () –,
78
on generalizing from Egypt. Sample size is households.
PBrux , .
79
In a stationary population about per cent of all families will have no natural heir, and a further
per cent will have a daughter or daughters only, irrespective of the prevailing mortality level.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Arsinoe, the couple has living with them not only their own daughter, but
an adult son and daughter from the husband’s two prior marriages (these
half-siblings have married), plus the wife’s son and daughter from her prior
marriage.80
In short, the most salient characteristics of Roman households are likely
to have been, ironically, their irregularity and unpredictability over time;
familial dislocation was a more or less inevitable aspect of life, and many
Roman social and legal institutions reflect the attempt to plan for such
vagaries.81 From this perspective, it is misleading to think in terms of a
typical ‘life cycle’ of the Roman family.
. Reproduction among slaves
For Roman demography, the institution of slavery presents a special
problem. By law, the offspring of a slave woman were slaves themselves,
the property of her owner; and birth is accordingly recognized as one
means whereby people became slaves. But the relative importance of birth
as a source of slaves is not easy to assess. Slaves could not marry, and
Roman law also did not recognize the legitimacy of informal slave unions
(contubernia); but such unions did receive at least a degree of social recognition and respect, and they are often mentioned in inscriptions. Still, it
remains deeply problematic whether, in general, the slave population
reproduced itself naturally.82
In Roman Egypt, slave families are rarely detectable in our sources; yet
the census returns, where slave mothers are usually listed with their children, show that masters expected their adult female slaves to bear children,
and that female slaves often did so. For this reason, female slaves were not
commonly manumitted while still of childbearing age. Rough statistics
indicate that slave fertility was probably about the same as that of all free
women, though well below that of free married women. 83
.
The third principal demographic component is migration. The early
empire erected few formal barriers to the movement of population within
its borders, and it is clear from literary works (such as the Satyricon) and from
inscriptions that mobile segments of the population took advantage of this
fact. Further, to some extent the empire itself remained receptive to outsiders, although their immigration was often involuntary. During the first
80
81
82
83
PTebt (= Sel Pap ), census of .. /.
Bradley () –; Saller (); Krause ().
See Harris (b) –; Bradley (). On contubernium: Treggiari (b).
Bagnall and Frier () –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
two centuries .., migration probably worked to the general advantage of
the western Mediterranean. To be sure, the numbers involved were
undoubtedly small in the short run; the huge mass of the population
remained attached by tradition, though not yet by law, to the soil of its
birth. Yet the balance between mortality and fertility was so delicate that in
the long run the transfer of even small numbers could have significant
effects.
. Within the empire
If we leave to one side the geographic movement of the political élite to
and from Rome, there were three main avenues of internal migration
during the early empire. The first was military colonization.84 After Actium,
Augustus had undertaken extensive resettlement of his veterans in strategically situated colonies throughout the empire. His successors continued
these efforts, though at a reduced pace and with the more specific aim of
settling empty or undercultivated areas; accordingly, the great majority of
new post-Augustan colonies were in the West, and in some areas, such as
Africa, they played a major role in urbanization.85 However, veterans generally disliked such enforced colonization, and under Hadrian the programme lapsed; veterans now settled mainly in the region of their service
or retirement. Nonetheless, periodic movements of troops still effected
some migration; a good and unusually well-documented example is the
Syrian cohort that in took up residence in Pannonia on the Danube.86
The second major avenue of migration involved the empire’s commercial and intellectual classes. Inscriptions make clear their drift towards the
western Mediterranean. For example, onomastic studies of African
inscriptions reveal large numbers of Syrians, Jews and other easterners who
were not yet assimilated;87 and similar evidence has emerged in Italy, Gaul
and Spain. Most of these immigrants were presumably drawn westward in
hope of gain, and they are accordingly concentrated especially in port
cities; but the wealth of the imperial capital was a powerful magnet also to
educated Greeks, as Lucian’s essay ‘On Salaried Posts in Great Houses’ candidly acknowledges. This network of easterners settled in the West was
crucial in spreading eastern religions, including Christianity.
The third major avenue of migration was the flow of slaves. Slaves are
a highly mobile form of capital, and it is not surprising to find evidence for
their widespread dispersion in all directions across the Mediterranean;
however, since slaves often received new names from their masters, onomastic studies of the slave trade are not easy. Literary sources indicate,
84
87
85
Mann Recruitment.
Lassère () –.
Lassère () –.
86
Fitz ().
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
usually in moralistic terms, that the great majority of household slaves in
Rome were of ‘eastern’ origin (usually Anatolians or Syrians); and in fact
the numerous epitaphs from Rome record about twice as many freedmen
(most vaguely ‘eastern’) as freeborn citizens.88 The Orontes was flowing
into the Tiber, so Juvenal put it. A similar but less pronounced pattern is
found throughout Italy and the western provinces, especially in port cities.
To be sure, epigraphic evidence may significantly overrepresent the proportion of freedmen in the general population; but heavy migration of
slaves from East to West is in any case evident, and it undoubtedly overwhelmed migration in the opposite direction.89
. Across the empire’s borders
The imperial army, stationed chiefly along the Rhine–Danube frontier to
the north-east and on the Armenian and Parthian frontier to the east,
engaged in more or less continuous skirmishes; since enslavement was the
usual fate of prisoners, it is not unlikely that –, slaves per year
entered the empire by this route.90 Major wars dramatically increased the
number of military slaves. An eyewitness states that Trajan’s Dacian campaigns yielded more than half a million slaves; this is doubtless exaggerated, but we have good evidence that the suppression of the Jewish revolt
in produced , slaves, and in the defeat of Simon Bar Kochba’s
revolt caused a glut in local slave markets.91 Further, Roman slave traders
continued to operate across the frontiers; their efforts perhaps equalled the
normal influx of slaves from military sources. A majority of these slaves
doubtless ended up in the West.
Augustus made fairly extensive efforts to populate the area inside the
north-eastern frontier by resettling large groups of tribesmen from beyond
it.92 Later Julio-Claudians continued the policy fitfully; in Claudius
allowed a Suebian king and his dependants to settle in Pannonia, and in
a governor of Moesia resettled within his province at least ,
Germans from across the Danube.93 Thereafter the practice apparently
lapsed for a century, until in about Marcus Aurelius revived it by resettling the Marcomanni.94 The emperor may have been motivated by depopulation following the plague of ; in the later empire resettlement was a
frequent imperial response to underpopulation.
Finally, some thinly settled areas at the empire’s periphery were annexed
after .. : Britain, Nabataea and Dacia. None of these is likely to have
88
91
92
93
94
89
90
Taylor ().
Harris (b) –.
Harris (b) .
Lydus, Mag. .; Joseph. BJ .; Chron. Pasc. . Dindorf.
Strab. .. (p. ), .. (p. ); Suet. Aug. .; Tac. Ann. ...
Tac. Ann. ..; ILS . See also Ste Croix, Class Struggle –.
Dio .., ; HA Marc. ., ..
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
had more than a few hundred thousand inhabitants at the time of annexation. Rome’s shifting border with Armenia and Parthia can be ignored for
present purposes.
.
The Roman demographic structure, though undeniably harsh by modern
standards, posed no obstacle to modest population growth. Most of
human history has been lived under conditions of mortality not unlike
Rome’s. Where peace, prosperity and freedom from general epidemic have
obtained, populations of the past have normally experienced a modest
measure of sustained growth; there is no reason to believe that the Roman
empire’s population did not grow similarly. The following section hypothetically reconstructs the pattern of such growth in the early empire.
. Regions and cities
The starting-point must be Julius Beloch’s famous attempt to estimate the
population of the empire and its regions in .. , the year of Augustus’
death.95 Beloch believed that the empire then contained about million
persons who were fairly evenly balanced between the Greek East, with
million inhabitants, and the Latin West, with million. The East, however,
was far more densely settled than the West.
By and large, Beloch’s prudent estimates have stood up extremely well
to subsequent criticism.96 The main difficulty is his estimate for Anatolia
and greater Syria, to which Beloch assigned a combined population of
million; this figure is incredible, since it requires a population density not
achieved again until the twentieth century. A figure of about million is
considerably more plausible.97 Otherwise, the likeliest modifications of
Beloch’s estimates have produced only a small cumulative downward effect
on his total. Table , based on one recent set of regional estimates for the
empire,98 suggests that in .. the total population was slightly more than
million persons, of whom about million resided in the East and
million in the West. These estimates imply that the entire Roman empire
had an average population density of . inhabitants per square kilometre,
obviously very low by modern standards; but the population density in the
95
Beloch () . Beloch’s later upward revision of these figures () is considerably less cred-
ible.
96
Summarized in Salmon () –. For a recent critique, see Lo Cascio (), with further bibliography. Aggrandizing estimates of ancient population often tacitly accept the hoary fallacy that more
is better; contrast the justified caution of Rathbone (), on Egypt.
97
Russell () –. Beloch’s estimates for Spain and Africa were also on the high side; cf. CharlesPicard () –, and Balil ().
98
McEvedy and Jones (); these estimates also closely resemble those of Russell ().
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Table : An estimate of the empire’s population in A.D.
Area
( km2)
Population
(millions)
Density
(per km2)
Greek East:
Greek peninsula
Anatolia
Greater Syria
Cyprus
Egypt
Libya
Total:
,.
,.
,.
,11.
,1.
,1.
,.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Latin West:
Italy
Sicily
Sardinia/Corsica
Maghreb
Iberia
Gaul/Germany
Danube Region
Total:
,.
,1.
,1.
,.
,.
,.
,.
,.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Roman empire:
,.
.
.
Sources: Land area: Beloch () . Population estimates: McEvedy and Jones (),
except that somewhat higher values are used for Anatolia, Greater Syria, Egypt and Italy.
Land areas include client kingdoms annexed soon after .. . The Greek peninsula
comprises the modern territories of Greece, Albania and Turkey in Europe. Greater Syria
includes Lebanon and Palestine.
East was almost twice that in the West. Only Italy and Sicily had achieved
a population density comparable to that generally obtaining in the East.
Since at this date annual gross national product per capita is likely to have
been about sesterces, the Roman Empire’s national product in ..
was about billion sesterces, equivalent in commodities to , tonnes
of gold. Comparative data from other pre-industrial economies suggest
that, of this national product, probably less than half was monetized and
about per cent derived from agriculture, in which at least three-quarters
of the workforce are likely to have been employed. Slaves may have constituted up to per cent of the empire’s population, thus as many as
million persons; but the proportion of slaves was considerably higher in
Italy, lower in Egypt and North Africa.99
99
Statistics in this paragraph derive from Goldsmith () –, except that a lower estimate of
total population is employed.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
The Roman empire was a network of cities; as many as a sixth of its
inhabitants resided in several thousand cities, a degree of urbanization that
is unusually high by pre-modern standards. By .. Rome, the imperial
capital, had at least , inhabitants, more than a tenth of Italy’s population, and in the following century it may have reached million; Rome
was larger than any Western city would be until the nineteenth century.100
The empire could support only one such city; other large cities numbered
only a few hundred thousand (Alexandria, Antioch, later Carthage), or less.
Most cities were much smaller, usually only –, persons of whom
many resided outside the city itself;101 still, the cumulative population of
these cities must have been at least – million. Further, small size gave
such cities a social stability that larger ones lacked; all the great cities of
antiquity, with their fetid conditions and high mortality rates, were heavy
net consumers of population.102
. Growth and decline
What happened to the Roman empire’s population after .. ? Clearly
there was room for growth, above all in the thinly settled West; and
although the population of Italy apparently remained stagnant, elsewhere
archaeology has provided conclusive evidence for growth, especially in
Africa, Spain and Gaul.103 Exact figures cannot be known, but table is an
attempt to suggest what the Roman population might have looked like in
.. . This table derives from a computer simulation using two reasonable assumptions. The first is that the population in both halves of the
Mediterranean grew at a fairly constant annual rate of . per cent during
the century and half from .. to ; this rate, though very low, would
double the population every . centuries.104 The second assumption is
that, on average, a net of , persons migrated each year from East to
West, and that , slaves also entered the empire each year, of whom
three-quarters ended up in the West. These rough calculations were periodically revised to reflect territorial annexations (though not the shifting
imperial border east of Syria), resettlement of barbarians within the
empire, and all major known disasters including natural catastrophes,
famines and plagues; the numbers obviously involve much guesswork, but
on the whole they have little impact on overall population figures.
In this simulation, the Roman empire’s population reached a peak of
about million persons in .. ; the Mediterranean basin would
not regain that level until the sixteenth century. The figure is obviously
hypothetical, but on any even remotely plausible assumptions the empire’s
100
101
102
Brunt () –.
Duncan-Jones, Economy –.
See just Scobie ().
Africa: Lassère () –; the three Gauls: Drinkwater () –; also Frere, Britannia
104
–, on Britain.
A rate of this order is supported by the Egyptian census returns.
103
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
Table : An estimate of the empire’s population in A.D.
Population
(millions)
Density
(per km2)
Increase from
.. (per cent)
Greek East:
Greek peninsula
Anatolia
Greater Syria
Cyprus
Egypt
Libya
Total:
Annexations
Total:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
—
.
.
.
Latin West:
Italy
Sicily
Sardinia/Corsica
Maghreb
Iberia
Gaul/Germany
Danube Region
Total:
Annexations
Total:
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
—
—
.
.
.
.
.
Roman empire:
.
.
.
Roman Pop.
reattained in:
}
}
Source: Modern population: McEvedy and Jones ().
population reached million persons, and – million is a much likelier
estimate. Table represents, on a highly provisional basis, how this growth
in population may have been distributed.
Significant growth was probably confined to the major peripheral provinces of the western empire, which gradually assumed a density comparable to that of Italy and the Greek East. This steady shift westward in the
balance of the empire’s population was in this theory accomplished entirely
by migration; therefore, if it were assumed that the intrinsic growth rate in
the heavily populated East was lower than in the West, the shift in population would be still more pronounced. Even on the assumption of equal
intrinsic growth rates, the East’s population remained virtually stagnant,
growing only per cent in a century and a half; the West increased at a rate
more than three times as high, with positive implications for economic
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
growth as well.105 This shift westward in population, made possible by
Rome’s political unification of the Mediterranean, was doubtless the most
important and enduring demographic achievement of the Roman empire.
The empire’s rising population might eventually have provoked a rise in
the cost of basic foodstuffs, at any rate after land in the West came under
more intensive cultivation. In Egypt the rise is clear: the median price of
wheat rose by about half from the first to the second century ..106
Elsewhere in the empire the phenomenon is far less securely attested, but
seems to have been similar. Rising food prices, if (as is likely) they were not
offset by an equivalent rise in real wages, ought to have pauperized an
increasing portion of the populace, but there is little reliable evidence for
this, and in particular no clear evidence that pauperization was acute
enough to cause decreased nuptiality and fertility, the preventive
Malthusian checks on population growth. This happy situation endured for
one hundred and fifty years.
By .. the Roman empire probably embraced about a fifth of all
persons then living. In land area and population, the Roman empire all but
duplicated the Eastern Han empire in China, where a census of ..
registered nearly million persons.107 Ten years later a Roman ‘legation’,
probably comprised of merchants, arrived in the Han capital, symbolizing
the quickening tempo of Rome’s contacts with South Asia. Such contacts
may well have caused a spill-over between hitherto isolated ‘disease pools’,
a spill-over that in unleashed upon the Roman empire a dreadful
plague.108
Fifteen years later, as he lay dying, Marcus Aurelius directed his friends
to mourn not him, but rather the general pestilence and death. The demographic consequences of the plague should not be exaggerated, but were
clearly severe.109 Literary sources, among them eyewitnesses such as Galen
and Dio, attest the plague in Egypt, Syria, Asia Minor, Greece, Italy, Gaul
and Germany; they also stress its heavy toll on human life both in cities and
on the land, its persistence and recurrence, and the widespread famines and
that broke out in its wake. The disease, probably smallpox, ‘behaved as
infections are wont to do when they break in upon virgin populations that
entirely lack inherited or acquired resistances. Mortality, in other words, was
heavy.’110 As much as per cent of the empire’s total population may have
105
Duncan-Jones (). Table is very tentative; recent archaeological evidence suggests, for
instance, that the population of the Hellenistic East may have grown more than this model allows.
106
Duncan-Jones, Structure –. Real rentals and wages probably did not keep up: Muth ().
In pre-modern societies, per cent or more of average household income is usually spent on food,
so the price of grain is of crucial importance.
107
108
On world and Chinese population: Durand () , .
McNeill () –.
109
Gilliam (); Salmon () –. Economic effects: Duncan-Jones, Structure –.
110
McNeill () ; see also Hopkins () –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press
.
perished in the plague; in cities and military camps the percentage was
perhaps twice as high.111 The plague would thus have undone about half a
century of slow growth. The Roman empire was not dealt a mortal blow;
but this sudden population drop ushered in, or immensely complicated, a
host of social and economic problems, to cope with which a new dynasty
was ultimately required.
111
These estimates are from Littman () –. On Egypt, see Rathbone () –.
Cambridge Histories Online © Cambridge University Press, 2008
https://doi.org/10.1017/CHOL9780521263351.028 Published online by Cambridge University Press