Slavery in the Roman Empire
Numbers and Origins

John Madden

Though slavery was a prevailing feature of all Mediterranean countries in antiquity, the Romans had more slaves and depended more on them than any other people.

It is impossible, however, to put an accurate figure on the number of slaves owned by the Romans at any given period: for the early Empire with which we are concerned conditions varied from time to time and from place to place. Yet, some estimates for Rome, Italy, and the Empire are worth attempting. The largest numbers were of course in Italy and especially in the capital itself. In Rome there were great numbers in the imperial household and in the civil service - the normal staff on the aqueducts alone numbered 700 (Frontin. Aq. 116-7). Certain rich private individuals too had large numbers - as much for ostentation as for work (Sen. Ep.110.17). Pedanius Secundus, City Prefect in AD 61, kept 400 slaves (Tac. Ann. 14.43.4), Gaius Caecilius Isidorus, freedman of Gaius Caecilius, left 4116 in his will in 8 BC, while some owners had so many that a nomenclator had to be used to identify them (Pliny HN 33.135; 33.26). However, there is evidence to suggest that these cases were not typical - even for great houses. Sepulchral inscriptions for the rich noble gens the Statilii list a total of approximately 428 slaves and freedpersons from 40 BC to AD 65. When these figures are analysed, the number of slaves and freedpersons definitely owned by individual members of the gens is small, e.g. Statilius Taurus Sisenna (consul of AD 16) and his son had six, Statilius Taurus Corvinus (consul ordinarius of AD 45) had eight, and Statilia Messalina, wife of Nero, four or five. Seneca, a man of extraordinary wealth, believed he was travelling frugally when he had with him one cartload of slaves (most likely four or five) (Ep 87.2). References in Juvenal and the Scriptores Historiae Augustae suggest that many non-plebeian Romans had either no slave or merely one or two (Sat. 3.286; 9.64-67,142-7; S.H.A. Hadr.17.6). From evidence such as this Westermann, Hopkins and others are understandably cautious when attempting to come to a total figure for slaves in the city of Rome in the 1st century AD. Hopkins' estimate of 300,000-350,000 out of a population of about 900,000-950,000 at the time of Augustus seems plausible.

The same kind of caution needs to be exercised in attempting to arrive at a figure for slaves in Italy for the same period. Passages in the Satyricon (e.g. 37;47;53) would suggest that some households had vast numbers. But that work is of course fiction - though the references to slave numbers there can only have point if certain private individuals did own a lot of slaves. Overall, a figure of around two million slaves out of a population of about six million at the time of Augustus would perhaps seem right (again we follow Hopkins). If so, approximately one in every three persons in Rome and Italy was a slave.

And what of the Empire as a whole for this period? It is impossible to give any kind of accurate figure. We have neither statistics for the total area nor for the provinces separately. And of course the number of slaves in each province depended on the particular circumstances prevailing there. Some provincial locations had a high number of slaves: Pergamum in the 2nd century AD (we deduce this from Galen De Propr. Anim. 9) had 40,000 adult slaves and these formed (as at Rome) one third of the adult population. At Oea (Tripoli) in Africa also in the 2nd century AD the wife of Apuleius owned a familia of slaves well in excess of four hundred (Apol. 77.93; cf.102). However, other areas in the Empire had comparatively few slaves. The evidence from papyri suggests that in all likelihood slaves in Egypt never rose much above 10% of the population and in poorer areas there dropped to as low as 2%. And in other regions, particularly perhaps in the more backward provinces of the West, slaves may never have comprised a significant segment of the work force at all. What then might we assume as an approximate number of slaves in the entire empire in this period? The attractive hypothesis of Harris is ten million, i.e. 16.6%-20% of the estimated entire population of the Empire in the first century AD, i.e. one in every five or six persons would have been a slave. This of course is not a computation, merely a conjecture.

Whence came these slaves? Some have presupposed that because two of the more important sources of slaves in the Republic - war and piracy - had become significantly restricted in the Empire there was a gradual diminution in the number of slaves during the first three centuries AD. However, there is no statistical proof of this, and for that reason Harris rejects it (rightly I believe), preferring to think that there was no serious drop in the number of slaves or in the demand for them - at least until AD 150. And since there is no evidence either that the cost of slaves spiralled upwards during this period, it seems sensible to infer that the supply of slaves needed annually to replenish the normal depletion of their numbers was more or less available without too much difficulty.

This raises two obvious but interesting questions:

1. What number of new slaves was needed from year to year?

2. Where did these replacements come from?

To answer the first of these questions we need to know the average length of time the slave spent as slave. This however, depends in turn on the average life expectancy of a slave. It has been estimated by Keith Hopkins that for the entire population the average life expectancy at birth was 20-30 years. Combining this figure with evidence from Roman tombstones Harris in turn estimates that the average life expectancy of a slave at birth in the Empire was unlikely to be more than 20 years. This seems reasonable, and since the average length of a slave's time as slave would be shorter than his average life expectancy at birth - partly because some slaves were manumitted and partly because some were not initially slaves but made so subsequently -it follows, if the number of slaves was to remain more or less constant as we have assumed, that the need for new slaves was exceptionally great. On these hypotheses Harris suggests that more than half a million were required annually for the first century and a half of the Empire.

Where did these slaves come from? The jurist gives a general answer: servi aut nascuntur, aut fiunt ['slaves are either born or made'] (Inst. 1.3). During the Republican period one of the principal sources (if not the principal source) of slaves had been prisoners of war. However, with the establishment of the principate under Augustus and the extension of the pax Romana across the Empire the significance of this source decreased. Yet, not completely of course: wars still continued but on a smaller scale. And there were even some major influxes of slaves from this source. The number of Jews enslaved as a result of the crushing of the Jewish rebellion by Vespasian and Titus (AD 66-70) was put (reliably, it would seem) at 97,000 by Josephus (B.J. 6.9.3). The steady expansion in Britain continued to supply British slaves onto the market. Great numbers of prisoners of war reached Rome from the Dacian wars of Trajan (John Lydus' half-a-million, [De Mag. 2.28] is, however, an exaggeration). And after the Jewish revolt led by Bar-Cochba in AD 132-35 a large amount of Jews - well over 100,000, it is estimated - were sold as slaves in the East (Chronicon Paschale 1.474).

Roman soldiers involved in frontier wars and rebellions would have had many chances to buy prisoners of war as slaves at disposal auctions. Although this is not mentioned in the contemporary literature, it can be deduced from papyri which reveal slaves in the ownership of soldiers and veterans in Egypt. However, when Hadrian decided on a border plan of continuous defence along natural or man-made boundaries, these opportunities must have become far fewer. The effect on slave numbers of these various military episodes though significant was yet more short than long term. Harris, for example, thinks it improbable that in an average year for the period AD 14-150 more than 2%-3% (i.e. 10,000-15,000) of the slave requirement was supplied from prisoners of war.

Accordingly, we must turn our attention to the other sources of slaves in the early centuries of the Empire. Some (e.g. Barrow, Boese) have taken the view that, since the slave body at this period was already very great, the bulk of new slaves required each year would have been provided from their own class, i.e. that the slave body would have been almost self-propagating. Vernae (i.e. slaves born at home and kept within the familia - in Roman law any infant born to a slave woman was in turn a slave [cf. Gaius, Inst. 1.82]) - are certainly mentioned frequently in our sources. They were normally preferred by the Romans, who tended to get on well with them: their background was known, they spoke Latin from the beginning, they were accustomed to slavery knowing no other life, and they could be taught whatever skill their master intended for them. In particular, we have indications that they were encouraged to marry and have children (cf. Columella, Rust. 1.8.19 and earlier Varro, Rust. 1.17.5,7; 2.126), and in fact for our period the slave's type of marriage - contubernium - is well documented. Surely then, the argument goes, the number of new slaves needed annually would come for the most part from reproduction among the slaves themselves?

However, on closer analysis, this reasoning is flawed. True, some of the more fortunate city slaves and certain rural ones as well enjoyed a secure home life. And undoubtedly these together with the many female slaves who had children by their masters (or other free men) will have contributed considerably to the number of new slaves entering the system each year. Nevertheless, the belief that the total slave-body was more or less self-propagating is unsound. There are a number of reasons for this.

One is the likelihood that the slave-body was disproportionately male. There is of course no clear statistical confirmation of this. However, if we allow for differences from one area to another and exclude entirely perhaps Roman Egypt (Biezunska-Malowist [1977], p.21, claims that male slaves were not more numerous than females there) the overall picture from the accessible evidence seems consistent.

1. First of all it is clear that males were in the majority where work was difficult and weighty - in building, in mining, in numerous types of industry, in a wide variety of services such as loading and unloading at docks, portage, transportation, etc. In agriculture also male slaves would have been more in demand. Small landowners would have to be content with whatever slaves were available irrespective of their sex, while large landowners would undoubtedly have needed some female slaves e.g. for weaving, cloth making, cooking. However, it is clear from passages in Varro and Columella, where the question of which of the more reliable agricultural slaves should be allowed a female companion is treated, that permission for such a partner was a special concession. Varro recommended that praefecti ['overseers'], as an incentive to their faithfulness, should be granted female slaves with whom they could have children, while lesser slaves should have to do with less (Rust. 1.17.5,7). In Columella, on the other hand, it is the vilicus ['steward'] who should be given a female partner (Rust. 1.8.4). In the ergastula - the private prisons belonging to many Roman farms where slaves were forced to work in fetters - the inmates would have been very largely male. It is evident from this that among agricultural slaves males surely outnumbered females.

When we turn to domestic staff the evidence suggests that there too male slaves were more numerous. S. Treggiari in her analysis of the 79 members of the city household staff of Livia has noted that 77% were male (the percentage was similar among freedpersons and slaves). This is a very revealing figure since we would expect a domina to have a higher number of female staff than a dominus. And in her study of the city familiae of the Statilii and the Volusii Treggiari has shown that about 66% of the freedpersons and slaves were male, while of the thirty child slaves whose names were inscribed on the tombs of these two families 80% at a minimum were male. P.R.C. Weaver in his examination of the burial places of the imperial household stationed in Carthage calculated that 76% were males. One of the Oxyrhynchus papyri provides evidence of a big urban slave familia in Roman Egypt. It belonged to the wealthy Titus Julius Theon in Alexandria (died AD 111) and of fifty-nine slaves (at least) recorded as belonging to it a mere two were female.

In literature also there are occasional indications that it was more usual for private individuals to have male rather than female slaves : e.g. Horace, Sat. 1.6.116, has his meal served "by three servant boys" (pueris tribus); Naevolus in Juvenal, Sat. 9.64-7, owns "one servant boy" (puer unicus) and will have to get a second; in Lucian, Merc. Cond. (=On Salaried Posts in Great Houses) 32 both the cook and lady's hairdresser seem to be male; in the Cena of the Satyricon male slaves appear to be almost everywhere, c.f. e.g.. 27;28; 31;34. It would seem safe to conclude that in general, whenever slaves were bought, males outnumbered females, and that this was the pattern also for the total slave body of the Empire. Incidentally, there are, as Harris points out, parallels for this from the later Atlantic slave trade. For the period 1791-1798 there were 230 male slaves brought to Cuba compared to 100 female. And for Jamaica for the same period the ratio was 183 males to 100 females.

A second reason for the slave body's inability to propagate itself is linked to manumission. There is some evidence to suggest that female slaves were manumitted more often than males and marriageable females (it would seem) most often of all. The principal reason for this is thought to be marriage. Since the children of such a marriage were usually free, the emancipation of these nubile slaves would have removed the very individuals who were essential for the continuation of the class.

Another reason for thinking that the slave population did not reproduce itself in sufficient numbers is that female slaves, on the available evidence (limited admittedly), do not seem to have been very prolific. Columella e.g. who refers to favours (such as a break from work) which he has granted to feminae fecundiores ['more fertile women'], includes mothers of three in this category. The implication is that in Italy the greater proportion of slave mothers gave birth to and reared at most two children. For the provinces unfortunately there is little reliable information. Even in Egypt the picture is obscure: the papyri do not seem to have been properly analysed with this question in mind. However, three or more children would appear to have been well above average for a slave woman there. P.Mich. V.326 mentions one slave mother of five children, but that is quite unusual. This apparently low fertility rate for Egypt is significant. There slaves brought in from abroad must have cost comparatively more than in the majority of other places across the empire. We would accordingly have expected even greater encouragement to procreation among slaves in Egypt than elsewhere.

Another reason for doubting that the slave body was likely to reproduce itself is that even free peoples at certain periods found it hard to do so. Polybius e.g. discusses the decline of the population of Greece in his own day (mid 2nd century BC) and attributes this to two factors, childlessness and the scarcity of men (36.17.5-11). Much of southern Italy also experienced a decrease in its numbers towards the end of the Republic. Indeed, it is remarkable - given the prevalence of child-exposure, the limitations of obstetrics, diet and medicine, and the effects of wars, famines and the like - how buoyant the numbers of free peoples in general remained.

The evidence from slave populations elsewhere in history perhaps deserves mention here. In general this favours the argument that slave populations do not normally reproduce themselves. In the United States the reverse was true (there slave numbers increased after importation was banned in 1808), but that was a special case. For example, Brazil and the Caribbean imported more African slaves than North America, yet the slave body in both areas experienced a natural decline in numbers - up to 5% per annum depending on time and place. Why the United States was different is not clear, but a plausible suggestion is that the working and environmental conditions affecting the lives of slaves were more favourable there than elsewhere. It is unlikely that comparable conditions were to be found in the Roman Empire, and so we would expect the trend there to resemble that which prevailed later outside the U.S.A.

If these arguments are correct, vernae on their own will have failed by a significant amount to meet the annual requirements for new slaves. This shortfall, Harris, taking everything into account, estimated at several hundred thousand per year.

How was this shortfall made good? Where did the required number of new slaves originate? We must turn again to those who were made slaves. As well as prisoners of war there were other groups who belonged to this category. First, there were those who were kidnapped and sold into slavery. Both Augustus and Tiberius took measures against kidnapping (Suet. Aug. 32.1; Tib. 8.2; Columella Rust. 1.6.3). Augustus put a curb on grassatores ['bandits'] who used to capture travellers (both free and slave) and hand them over to landowners for retention in ergastula. However, in spite of the efforts of the two emperors kidnapping was not eradicated and persisted down to the next century (cf. e.g. Apul. Met. 7.9; Philostr. V.A. 8.7.12). Yet, it would not have been so frequent and widespread that it could be considered a major source of new slaves in the early empire.

The same can be said of piracy. This practice was considerably restricted when Pompey crushed the pirates after the passing of the Lex Gabinia in 67 BC and later when the moves against the piratical Illyrians came to a successful conclusion at the battle of Actium in 31 BC Thereafter, the major section of the Mediterranean was rendered safe for journeying and commerce for the following two centuries by the creation and continued upkeep of the imperial fleet. However, some instances of piracy still occurred in the Mediterranean, cf. e.g. Lucian De Merc. Cond. 24. Outside the Mediterranean, where the auxiliary imperial squadrons were less active, pirates would have had more opportunities, cf. e.g. Strabo 11.496 for the Heniochi in the Black Sea. Nevertheless, it is improbable that piracy could ever have contributed a sizeable fraction of the slave numbers needed in the early centuries of the empire.

Another source of slaves was purchase from over the boundaries of the empire. Presumably this happened at all major limites, but its extent is impossible to determine. M. Crawford (JRS [1977],117-124) has made an attractive case for a series of such purchases in the late Republic. He argues that hoards of Republican denarii found in Dacia in the lower Danube basin are best explained as payments from c. 65 to 30 BC (i.e. after a hitherto significant source - piracy - had been suppressed by Pompey in 67 BC) by slave dealers/merchants to the local aristocracy. These latter (Crawford suggests) in exchange for sought-after commodities of the Mediterranean world such as silver and wine, sold off "perhaps [their] own humble dependants and certainly the humble dependants of others captured in internal raiding"(p.121).

For the Empire Tacitus gives an instance for the lower Rhine in the reign of Domitian (Agr. 28.5). Evidence for the practice can also be extrapolated from the Periplous Maris Erythraei (13,31,36), Strabo (11.493) and two tariff inscriptions, one from Palmyra (AD 137) (CIS II.3.1.3913), the other from Numidia (AD 202) (CIL VIII 4508). Yet, it is remarkable that when the place of origin of slaves is indicated in Roman literature, this is almost invariably from within rather than from without the Empire. And the papyri and inscriptions suggest the same pattern. Thus it is unlikely that slaves bought from beyond the frontiers would have been plentiful enough to meet a significant portion of the total slave needs of the early Empire.

What of the other sources of slaves? The sale of their own offspring by parents was one of these. This occurred particularly in hard times when parents attempted to ease their burden. The evidence for it, whether in literature or inscriptions, is sparse, yet we can presume that it did take place during the first centuries of the empire. However, the practice is unlikely to have been widespread. Philostratus says that it was a custom for the Phrygians "even to sell their children"(V.A. 8.7.12) - the inference being that this was rather exceptional. Tacitus (Ann. 4.72) tells how the Frisians in Lower Germany on being subjected to an excessive tribute by the Romans were forced eventually to sell their wives and children into slavery. This too however, would have been unusual. In general it is unlikely that even the most impoverished parents, once they had initially resolved to bring up a baby, would sell that baby into servitude - unless there was some very special provocation.

A few other methods of enslavement should also be mentioned here. The first was self-sale. Hermeros, for example (if we can take Petronius seriously in the Satyricon 57.4), rather than remain a tribute-paying provincial and hoping subsequently to become a Roman citizen (i.e. tribute-free), seems to have sold himself into slavery. A second method was for debt. Here a debtor who was unable to pay could be "given up" (addictus ) to his creditor. A third method was penal enslavement i.e. slavery arising from conviction in law. Punishment for grave crimes could entail the removal of personal rights - the guilty being usually condemned to work as slaves in quarries or mines or as gladiators. However, it seems unlikely that any one of these methods, or indeed all three together, could have provided a significant number of slaves for our period. In fact, in the case of self-sale, its legality was never formally acknowledged in Roman law.

We can now pause here and take stock. Of all the sources so far discussed vernae were the most important. But even they, it has been suggested, left on average an annual deficit of several hundred thousand slaves. Since none of the other sources mentioned made significant dents in this number, we must ask "From where then did the main bulk of the remainder come?" Evidence points to "foundlings" - a source which, we may surmise, supplied considerably more slaves for our period than had been usually thought.

The abandonment of infants was widespread over much of the Roman world, and, no doubt, occurred even more frequently whenever circumstances became especially difficult. The custom was not made illegal until AD 374. Abandoned children usually either died or were made slaves, but the percentage in each group is beyond recall. In the case of the latter the owners themselves sometimes found the infants (either by accident of design); at other times they received them from finders who knew of their need. But there are also signs in the papyri of the availability of infants on request i.e. that individuals who were part of the slave trade nexus either collected abandoned babies for later sale themselves or bought them from others who found them. Sometimes owners engaged nurses under contract (a number of the relevant documents survive among the papyri) to look after foundlings in their early months. Occasionally foundlings were recovered by their parents. At other times if they could provide proof of their original citizen status and obtain a champion - an assertor - they could themselves initiate legal proceedings.

Evidence for the practice of child-exposure in Rome and across the Empire is considerable. In his Roman Antiquities 2.15 Dionysius of Halicarnassus refers to a law of Romulus which obligated the Romans to raise all infant boys and the first-born infant girl, and prohibited the killing of any child under the age of three unless it was maimed or abnormal from birth. The likelihood is that this was not a real law but an invention in the late Republic in response to the worrying level of child-abandonment at the time (v. P.A. Brunt, Italian Manpower, 149). Tacitus considered it deserving of comment (Germ. 19; Hist. 5.5) that the Germans and the Jews held it wrong to kill an unwanted child - the implication being that the Romans thought otherwise. Even the Roman aristocracy exposed children. Dio Cassius (54.16) says that there were far more males than females among the nobility in Rome (in 18 BC), a ratio that can best be explained (given the number of males killed in the civil wars) by the abandonment of females at birth. This hypothesis receives support in Pliny, Pan. 26.5-27, where it is clear that not only did the poor need incentives to bring up their children but the rich did as well. Augustus would not allow the child born to Julia the Younger to be acknowledged and reared (Suet. Aug. 65).

Elsewhere in many areas of the Roman world, we find child exposure widespread. It is securely documented for Roman Egypt, where two graphic phrases from the papyri deserve mention. The first is the legal description of the practice: " to rescue from the dung-pile for enslavement" (e.g. B.G.U. 4.1107.9; 5.41); the second, the infamous advice of a husband in a letter to his wife in 1 BC :"If you do give birth, if it is male, let it live, if it is female, cast it out" (Select Papyri [Loeb] 1 no.105,8-10). (Needless to say, not all Egyptian-Greek couples thought like the husband here: infant boys too were abandoned there.)

Child-exposure was practised in Asia Minor, on the Greek mainland and on the Aegean islands. In Bithynia-Pontus for example in AD 111 during the governorship of Pliny the Younger the problem of the status and maintenance costs of enslaved foundlings became so serious that Pliny felt compelled to write to Trajan himself about the matter - and received from the emperor a letter in reply (Ep. 10.65-66). Pliny's letter also reveals that similar kinds of problems had long caused concern in Achaea, for it refers to an edict of Augustus, and to letters of Vespasian, Titus and Domitian dealing with these matters there. Aelian, V.H. 2.7, thought it noteworthy that the citizens of Thebes attempted to put a stop to child-exposure. Plutarch, De amore prolis 5 (Mor. 497e), stated bluntly: "poor people do not rear their children". Christian apologetic writers too (even though there may be exaggeration in their censure ) indicate that the practice remained widespread: cf. e.g. Tertullian, Apol. 9.7, "you (i.e. pagans) in more cruel fashion stifle your children's breath in water, or expose them to cold and hunger and dogs"; see also idem , Ad Nat. 1.15; Minucius Felix, Oct. 30.2; Justin Martyr, Apol. 1.27.

Granted all of this, how important are we to rate foundlings as a slave source? Here we may recapitulate and conclude. When Tiberius and his successors followed in general Augustus' advice of confining the empire within its present frontiers (Tac. Ann. 1.11.7), one of the principal sources of slaves i.e. prisoners of war, seriously decreased. However, as far as we are aware, no major emergency in the replenishment of slave numbers occurred. The reason surely is that the considerable range of other slave sources already available were together able to make up sufficiently for the shortfall. Yet, of these sources, only two, vernae and foundlings, could, we are convinced, have been major contributors. Vernae certainly were important, yet, if the arguments and estimates given above are sound, vernae would have fallen well short of supplying the full yearly requirement. That leaves foundlings. Since we know that child exposure was a widespread phenomenon - urban as well as rural - over a considerable section of the Greek world, Asia Minor, Egypt, Italy and elsewhere in the West, we may conclude that enslaved foundlings were more or less able to make up for the shortage in the slave supply caused by the drop in the numbers of prisoners of war in the first century and a half of the empire - a shortage which vernae and the other sources could not offset.

Additional Bibliography

Biezunska-Malowist, I. "Die Expositio von Kindern als Quelle der Sklavenbeschaffung im griechisch-romischen Aegypten". Jahrbuch für Wirtschaftsgeschichte 2 (1971) 129-33.

__________________ L'esclavage dans l'Egypte gréco-romaine. Seconde

partie:période romaine. Warsaw,1977.

Boese, W. E. A Study of the Slave Trade and the Sources of Slaves in the Roman Republic and the Early Roman Empire . University of Washington: Dissertation, 1973 [obtainable in book form from University Microfilms, Ann Arbor, Michigan].

Bradley, K.R. "On the Roman Slave Supply and Slavebreeding". In M.I.Finlay, ed. Classical Slavery . London, 1987. [This article emphasizes the "high probability" ofvernae as an important source throughout Rome's central period (i.e. c.200 BC-c.AD200), but argues against regarding "warfare under the Republic and breeding under the Empire as the two respective principal sources of supply," preferring rather "a combination of complementary sources, warfare, trade and breeding, each of which was a constant contributor but of fluctuating relative significance according to developments in the wider historical world"(pp.58-9).]

__________ Slavery and Society at Rome . Cambridge,1994.

Crawford, M. "Republican Denarii in Romania: the Suppression of Piracy and the Slave-Trade". JRS 67 (1977), 117-24.

Harris, W.V. "Child-Exposure in the Roman Empire". JHS 84 (1994), 1-22.

Treggiari, S. "Jobs in the Household of Livia". Proceedings of the British School in Rome 43 (1975), 48-77.

__________"Family Life among the Staff of the Volusii". TAPA 105 (1975), 393-401

Weaver, P.R.C. Familia Caesaris. Cambridge,1972.

Wiedemann,T.E.J. Slavery . Greece and Rome : New Surveys in the Classics No. 19. Oxford,1987.

Note