Hostname: page-component-586b7cd67f-l7hp2 Total loading time: 0 Render date: 2024-11-25T08:25:29.194Z Has data issue: false hasContentIssue false
  • English
  • Français

Roman Age Structure: Evidence and Models*

Published online by Cambridge University Press:  14 March 2012

Walter Scheidel
Walter Scheidel*
Affiliation:
University of Chicago
*
[email protected]
Get access Rights & Permissions [Opens in a new window]

Extract

For many Romans, life was short. In consequence, the young greatly outnumbered the elderly. Historians have long accepted these basic truths, even if they are only beginning to come to terms with the social implications of an alien demographic regime. But how short is ‘short’, and how many Romans were children, how many adults? Does it matter, and can we know?

The importance of demographic structure is not in doubt. High mortality causes scarce energy resources to be wasted in pregnancies and nursing, and poses a disincentive to investment in education. It destabilizes families and households, exposes orphans and widows to risk and potential hardship, and shortens the time-horizons of economic activity. In the long term, average life expectancy is the principal determinant of fertility. Poor chances of survival trigger high birth rates to ensure genetic survival. High fertility, in turn, is negatively correlated with the status and well-being of women, and constrains female participation in economic and public affairs. Overall age structure, in conjunction with cultural practices from marriage to child care, determines the prevalence of orphans and widows, and affects the age-specific distribution of fertility. In sum, age structure is instrumental in framing and shaping expectations and experiences. For this reason alone, our understanding of life in the Roman world is critically dependent on our knowledge of demographic conditions.

Type
Articles
Information
The Journal of Roman Studies , Volume 91 , November 2001 , pp. 1 - 26
Copyright
Copyright © Walter Scheidel2001. Exclusive Licence to Publish: The Society for the Promotion of Roman Studies

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Footnotes

*

This paper is meant as a reprise of Keith Hopkins' critique of earlier attempts to explore Roman age structure (see below, n. 6). Hopkins' pioneering study of 1966 encouraged a renaissance of Roman population studies without which contributions such as my own would have been impossible. His retirement as Professor of Ancient History in the University of Cambridge provides a fitting occasion for my dedication. I wish to thank Peter Garnsey, Richard Saller, and the journal's referees for comments on an earlier draft of this paper.

References

1 For some first attempts, see Saller, R. P., Patriarchy, Property and Death in the Roman Family (1994)CrossRefGoogle Scholar; ‘Legal institutions, population structure and investment in the Roman Empire’, paper delivered at the Conference on Institutions and Markets in Comparative-Historical Perspective (Stanford, 1999)Google Scholar.

2 The Roman kinship simulations in Saller, op. cit. (n. 1), 43–69, seek to capture the essence of Roman domestic experience. For patterns of fertility, see Frier, B. W., ‘Natural fertility and family limitation in Roman marriage’, CP 89 (1994), 318–33Google ScholarPubMed.

3 Figs 1–2: A. J. Coale and P. Demeny, Regional Model Life Tables and Stable Populations (1983), 57, 79.

4 Coale and Demeny, op. cit. (n. 3), 4.

5 I omit reference to straightforward readings of the extant sources as a faithful mirror of reality, which lack intellectual warranty and have deservedly been in decline since the 1960s.

6 This approach was first advocated by Hopkins, K., ‘On the probable age structure of the Roman population’, Population Studies 20 (1966), 245–64,CrossRefGoogle ScholarPubMed re-iterated in his ‘Graveyards for historians’, in F. Hinard (ed.), La mort, les morts et I'au-delà dans le monde romain (1987), 113–26, and endorsed by T. G. Parkin, Demography and Roman Society (1994), chs 1–2. Cf. also Saller, op. cit. (n. 1), 22–3. For the nature of model life tables, see below, Section II.

7 Frier, B. W., ‘Roman life expectancy: Ulpian's evidence’, HSCP 86 (1982), 213–51;Google ScholarPubMed‘Roman life expectancy: the Pannonian evidence’, Phoenix 37 (1983), 328–44;CrossRefGoogle Scholar R. S. Bagnall and B. W. Frier, The Demography of Roman Egypt (1994), chs 4–5.

8 Frier, B. W., ‘Demography’, in CAH 2 XI (2000), 787–816,Google Scholar esp. 788–97.

9 Above n. 3.

10 A ‘stable population’ is a (hypothetical) population with fixed birth and death rates that will therefore grow or shrink at an unchanging rate (or not change at all). See, e.g., C. Newell, Methods and Models in Demography (1988), 120–2.

11 Coale and Demeny, op. cit. (n. 3), 12, 24–5.

12 Coale and Demeny ‘exclude the possibility of using these life tables for the purpose of generalizing on mortality conditions unless the validity of each table is carefully analyzed’ (op. cit. (n. 3), 4). However, as long as life tables based on more highly developed populations provide the only practical standard of reference, the circular character of any such ‘analysis’ is obvious.

13 op. cit. (n. 3), 25. Model life tables for non-Western countries draw on recent sources that resemble Western data: e.g., Unabridged Model Life Tables Corresponding to the New United Nations Model Life Tables for Developing Countries (1982). For more unconventional attemps to exploit non-Western material, see S. Ledermann, Nouvelles tables-types de mortalité (1969); L. Petrioli, Nouvelles tables-types de mortalité (1982).

14 Coale and Demeny, op. cit. (n. 3), 5.

15 M. Livi-Bacci, A Concise History of World Population (1992), 108.

16 P. A. Brunt, Italian Manpower 225 B.C.–A.D. 14 (1971, repr. 1987), ch. 5.

17 Athens in the Peloponnesian War is a case in point: M. H. Hansen, Three Studies in Athenian Demography (1988), 14–28. L. H. Keeley, War Before Civilization (1996), 88–94, illustrates the huge demographic impact of war deaths in small pre-state societies.

18 P. Garnsey, Famine and Food Supply in the Graeco-Roman World (1988), 6–39. For the comparatively moderate demographic impact even of genuine famines, see Watkins, S. C. and Menken, J., ‘Famines in historical perspective’, Population and Development Review 11 (1985), 647–75CrossRefGoogle Scholar.

19 Duncan-Jones, R. P., ‘The impact of the Antonine plague’, JRA 9 (1996), 109–11Google Scholar.

20 D. Herlihy and C. Klapisch-Zuber, Tuscans and their Families: a Study of the Florentine Catasto of 1427 (1985), 195.

21 A. W. Clark, ‘Mortality, fertility, and the status of women in India, 1881–1931’, in T. Dyson (ed.), India's Historical Demography (1989), 134–5.

22 Omran, A. R., ‘The epidemiological transition: a theory of the epidemiology of population change’, Milbank Memorial Fund Quarterly 49 (1971), 509–38,CrossRefGoogle ScholarPubMed esp. 516; Perrenoud, A., La population de Gèneve du seizième au début du dix-neuvième siècle I (1979), 429Google Scholar.

23 e.g., Howell, N., ‘Demographic anthropology’, Annual Review of Anthropology 15 (1986), 219:CrossRefGoogle ScholarPubMed ‘When the parameters are known (…), the population structure that will result is highly predictable and can be projected backward and forward in time to examine the implications of sets of parameters’.

24 Woods, R. I., ‘On the historical relationship between infant and adult mortality’, Population Studies 47 (1993). 195–219,CrossRefGoogle ScholarPubMed esp. 204–12.

25 E. A. Wrigley et al., English Population History from Family Reconstitution 1580–1837 (1997), 284, find that in England in the 1680s, mortality up to age fifteen corresponded to Coale/Demeny Model North Level 8 (implying a mean life expectancy at birth [e 0] of 36 years) while adult mortality was closest to Level 2 (e 0 = 21.2). Although they consider this a transitional period, it deserves notice that even at the end of that transition mortality up to age fifteen resembled Level 11 (e 0 = 43.4) while adult mortality was closer to Level 9 (e 0 = 38.4). In both cases, adult mortality relative to child mortality was higher than predicted.

26 P. N. Mari Bhat, ‘Mortality and fertility in India, 1881–1961: a reassessment’, in Dyson, op. cit. (n. 21), 73–118 (discussing China as well as India).

27 O. J. Benedictow, The Medieval Demographic System of the Nordic Countries (1993), 26.

28 R. M. Smith, ‘Demographic developments in rural England, 1300–48: a survey’, in B. M. S. Campbell (ed.), Before the Black Death (1991), 59.

29 Lee, J. Z., Wang, F. and Campbell, C., ‘Infant and child mortality among the Qing nobility: implications for two types of positive check’, Population Studies 48 (1994), 398, 401CrossRefGoogle Scholar.

30 e.g., M. J. Dobson, Contours of Death and Disease in Early Modern England (1997), 175–8.

31 Feeding and weaning: Garnsey, P., Cities, Peasants and Food in Classical Antiquity (ed. W. Scheidel) (1998), 261–70Google Scholar. Cereal food given forty days after birth (‘as do those for the most part <who> find nursing a burden’): Soran., Gyn. 2.46. Duration of wetnursing contracts: Masciadri, M. M. and O. Montevecchi, I contratti di baliatico (1984), 32–5Google Scholar. Isotope analysis of skeletal remains has the potential to illuminate differences in weaning practices: for a first attempt, see Prowse, T. L. et al. , ‘Chemical analysis of infant feeding practices from the imperial Roman site of Portus Romae, Italy’, American Journal of Physical Anthropology suppl. 30 (2000), 254Google Scholar.

32 e.g., Knodel, J. C., Demographic Behavior in the Past (1988), chs 3–4CrossRefGoogle Scholar. There are likewise great differences in the mean duration of breastfeeding (and thus birth intervals): e.g., Wood, J. W., Dynamics of Human Reproduction (1994), ch. 8Google Scholar.

33 e.g., Livi-Bacci, M. , Population and Nutrition (1991)CrossRefGoogle Scholar; Johansson, S. R., ‘Food for thought: rhetoric and reality in modern mortality history’, Historical Methods 27 (1994), 101–25CrossRefGoogle Scholar.

34 Retrospective medical diagnosis from inadequate data is a hazardous business. In the following, I will focus on major diseases whose distinctive symptoms are reasonably well described in pre-scientific and even ancient sources. In addition, skeletal data (from bone lesions to pathogen DNA) increasingly contribute to our knowledge of medical conditions in early societies. Grmek, M. D., Diseases in the Ancient Greek World (1989)Google Scholar provides a masterful account; for a state-of-the art survey of the physical evidence, see Aufderheide, A. C. and Rodríguez-Martin, C., The Cambridge Encyclopedia of Human Paleopathology (1998)Google Scholar. Dobson, op. cit. (n. 30), ch. 5, discusses more recent textual evidence; cf. also in general Alter, G. and Carmichael, A., ‘Studying causes of death in the past: problems and models’, Historical Methods 29 (1996), 44–8CrossRefGoogle Scholar.

35 e.g., Puranen, B., ‘The decline of mortality in Sweden’, in R. Schofield et al. (eds), The Decline of Mortality in Europe (1991)Google Scholar, 84 fig. 4; Woods, op. cit. (n. 24), 213 fig. 13.

36 Coale and Demeney, op. cit. (n. 3), 11–12.

37 Barclay, G. W. et al. , ‘A reassessment of the demography of traditional rural China’, Population Index 42 (1976)CrossRefGoogle Scholar, esp. 621. As a result, the Chinese mortality pattern is ‘different from the structure of any model life tables’ (624).

38 Grmek, op. cit. (n. 34), 183–94. For tuberculosis in ancient Egypt, see Scheidel, W., Death on the Nile: Disease and the Demography of Roman Egypt (in press), ch. 1.3.6.CrossRefGoogle Scholar

39 Dobson, op. cit. (n. 30), 133–220, esp. 168–72.

40 Del Panta, L., Malaria e regime demografico (1989), 22Google Scholar. Cf. also Arlacci, P., Mafia, Peasants and Great Estates (1983), 182 (Crotonese)Google Scholar.

41 Sallares, R., Malaria and Rome: a History of Malaria in Central Italy in Antiquity (in press)CrossRefGoogle Scholar. On the city of Rome, see also Scheidel, W., ‘Germs for Rome’, in C. Edwards and G. Woolf (eds), Rome the Cosmopolis (in press)Google Scholar.

42 Lee, Wang and Campbell, op. cit. (n. 29), 398 fig. 1; Coale and Demeney, op. cit. (n. 3), 42.

43 Rogers, L., Dysenteries (1913)Google Scholar, 114!–15, 284–5; Lancaster, H. O., Expectations of Life (1990), 74CrossRefGoogle Scholar.

44 Schnepp, M. B., ‘Considérations sur le mouvement de la population en Egypte’, Mémoires ou travaux originaux presentés et lus à I'Institut Egyptien 1 (1862), 552–3;Google ScholarPanzac, D., ‘Endémies, épidémies et population en Egypte au XIXe siècle’, in L'Egypte au XIXe siècle (1982), 89 tab. 7Google Scholar. See also below, Section IV.

45 Huckstep, R. L., Typhoid Fever and other Salmonella Infections (1962), 49Google Scholar; Hornick, R. B., ‘Typhoid fever’, in A. S. Evans and H. A. Feldman (eds), Bacterial Infections of Humans (1982), 660, 667Google Scholar; Lancaster, op. cit. (n. 43), 71.

46 For typhoid, see Grmek, op. cit. (n. 34), 89, 346–7.

47 Mittler, D. M. and van Gerven, D. P., ‘Developmental, diachronic, and demographic analysis of cribra orbitalia in the medieval Christian populations of Kulubnarti’, American Journal of Physical Anthropology 93 (1994), 287–97CrossRefGoogle ScholarPubMed.

48 Garnsey, , op. cit. (n. 31), 249, and Food and Society in Classical Antiquity (1999), 56–7CrossRefGoogle Scholar. Cf. Grmek, op. cit. (n. 34), 276, for Greece.

49 Lee, J. and Campbell, C., Fate and Fortune in Rural China (1997)CrossRefGoogle Scholar, 63 tab. 4.2. For instance, male mortality between ages twenty-one and thirty-five corresponded to Model North Level 12 Males but matched Levels 3 to 5 after age fifty-one. Among women, death rates in the late teens went right off the chart, implying a mortality level of minus 8 (i.e., eight levels worse than the most pessimistic model life table), but oscillated between Levels 2 and 7 for most other age brackets.

50 See Garnsey, op. cit. (n. 48), 100, for the probable effects on longevity.

51 Zhao, Z., ‘Long-term mortality patterns in Chinese history: evidence from a recorded clan population’, Population Studies 51 (1997), 117–27CrossRefGoogle ScholarPubMed. But contrast the Chinese data discussed above, n. 49.

52 For a similarly sceptical overview, see Parkin, op. cit. (n. 6), 41–58. For detailed bibliography covering the debate over the past twenty years, see W. Scheidel, ‘Progress and problems in Roman demography’, in W. Scheidel (ed.), Debating Roman Demography (2001), 19 n. 66.

53 Demographic analysis of epitaphs is now largely discredited, thanks to Hopkins, op. cit. (n. 6, 1966) (and again op. cit. (n. 6, 1987), 121–6); see also briefly R. Duncan-Jones, Structure and Scale in the Roman Economy (1990), 101–3, and in more detail Parkin, op. cit. (n. 6), 5–19. Scheidel, op. cit. (n. 52), 17 n. 59, provides further references. Clauss, M., ‘Probleme der Lebensalterstatistik aufgrund römischer Grabinschriften’, Chiron 3 (1973), 395–417,Google Scholar documents extreme local variation in implied age structures. For changing patterns of commemoration, see B. D. Shaw, ‘The cultural meaning of death: age and gender in the Roman family’, in D. I. Kertzer and R. P. Saller (eds), The Family in Italy from Antiquity to the Present (1991), 66–90.

54 e.g., Frier, op. cit. (n. 8), 790 (‘what is generally conceded to be by far the best surviving demographic source for ordinary subjects of the Roman empire’). Parkin, op. cit. (n. 6), 21, 59, is more sceptical. For a more detailed demographic re-appraisal of these documents, readers are referred to ch. 2 of my book cited above, n. 38.

55 Bagnall and Frier, op. cit. (n. 7); R. S. Bagnall, B. W. Frier and I. C. Rutherford, The Census Register P.Oxy. 984 (1997). Tax lists: Bagnall and Frier, 102–3.

56 All my calculations are based on the age data compiled by Bagnall and Frier, op. cit. (n. 7), 334–6, with Frier's corrections in Bagnall et al., op. cit. (n. 55), 113, and supplemented by thirty-seven additional references in Bagnall and Frier, 309–12; Nelson, C. A., ‘Four papyri from the Berlin collection’, BASP 32 (1995), 123–32;Google ScholarSijpestein, P. J., ‘Three papyri concerning census’, ZPE 107 (1995), 271–6;Google ScholarDuttenhöfer, R., ‘Five census returns in the Beinecke Library’, BASP 34 (1997), 53–78;Google Scholar P. Prag II 127.

57 Bagnall and Frier, op. cit. (n. 7), 75–110 (quote: 109). Cf. 34 n. 10 for a disclaimer concerning infant mortality.

58 ibid., 109 (Egypt), no (‘the returns provide solid support for the emerging picture of life expectancy in the early Roman empire’).

59 ibid., 106–9. ‘We are much less confident about our restored male life table [sc. than about the female table], although we believe that it may not be far off in approximating the true level of male mortality, particularly among adults’ (109).

60 The tax list pattern in Bagnall and Frier, op. cit. (n. 7), 103, is based on a garbled sample of evidence. For this graph, see Scheidel, op. cit. (n. 38), ch. 2.3.1.

61 The age data up to age sixty are broadly consistent with Model West Level 12 Males, with an implied mean life expectancy at birth of 44.5 years, a rate Egypt did not reach until the mid-twentieth century. This model is used for comparative purposes only.

62 For consideration of the possibility that they migrated to cities, see below, Section V.

63 Farris, W. W., Population, Disease, and Land in Early Japan, 645–900 (1985)Google Scholar, reporting low sex ratios similar to those found in Egyptian villages (Bagnall and Frier, op. cit. (n. 7), 163). Farris, 34, attributes this imbalance to attempted tax evasion. Just like in Egypt, this practice creates the specious impression of higher male life expectancy (ibid., 43; cf. Bagnall and Frier, 101).

64 T. F. Liao, paper given at the 25th Anniversary Meeting of the Social Science History Association (Pittsburgh, 2000).

65 Del Panta, op. cit. (n. 40), 22.

66 Dobson, op. cit. (n. 30), 148 fig. 3.18; cf. also 153.

67 Sallares, op. cit. (n. 41), drawing on Plin., Ep. 5.6.1, 6, 46.

68 I should stress that these examples should best be understood as demarcating opposite ends of a spectrum: not every malarial locale was as lethal as the Maremma, and some malaria-free areas could be unhealthy for other reasons.

69 For a judicious discussion, see esp. J. De Vries, European Urbanization 1500–1800 (1984), 175–98; more recent contributions include the articles in Annales de Démographie Historique 1990, 5–151, and Galley, C., ‘A model of early modern urban demography’, Economic History Review 48 (1995), 448–69CrossRefGoogle Scholar. For references to scholarship before 1984, see Scheidel, op. cit. (n. 52), 28 n. 106.

70 The classic analysis is Wrigley, E. A., ‘A simple model of London's importance in changing English society and economy 1650–1750’, Past and Present 37 (1967), 44–70CrossRefGoogle Scholar (various reprints). N. Morley, Metropolis and Hinterland (1996), 33–54, applies this model to the ancient city of Rome.

71 Scheidel, W., ‘Emperors, aristocrats and the Grim Reaper: towards a demographic profile of the Roman élite’, CQ 49 (1999), 254–81CrossRefGoogle ScholarPubMed.

72 e.g., Livi-Bacci, op. cit. (n. 33), 63–7; Johansson, op. cit. (n. 33), 113–14.

73 T. H. Hollingsworth, ‘Mortality in the British peerage families since 1600’, Population, numéro spécial (1977), 327–8; Lee, Wang and Campbell, op. cit. (n. 29), 401.

74 Provenance of census texts: Bagnall and Frier, op. cit. (n. 7), 8.

75 A charm from the Fayum village of Tebtunis targeted various types of malaria (PMG 33.1–25); for other ancient data, see, e.g., Cockburn, A., ‘Ancient parasites on the west bank of the Nile’, The Lancet 8252 (1981), 938;CrossRefGoogle ScholarMiller, R. et al. , ‘Diagnosis of Plasmodium falciparum infections in mummies using the rapid manual ParaSight™-F test’, Transactions of the Royal Society of Tropical Medicine and Hygiene 88 (1994), 31–2;CrossRefGoogle Scholar Hdt. 2.95. Thirteenth century: Nabulsi' description in Zeki, A., ‘Une description arabe du Fayoum au VIIe siècle de l'hégire’, Bulletin de la Société Khédivale de Géographie, ser. 5, 1 (1898), 264–9Google Scholar. Early twentieth century: Preliminary Report of the Anti-Malarial Commission (1919), 33–4; Report No. 1 of the Anti-Malarial Commission for the Period from 1919 to March 1925 (1928), 37–8; Report No. 13 on the Anti-Malarial Work in Egypt, 1936 (1939), 3. For a comprehensive discussion of malaria in Egypt from antiquity to the present, see Scheidel, op. cit. (n.38), ch. 1.3.5.

76 Annuaire statistique 1926–1927 (1928), 61. I am grateful to B. D. Shaw for providing me with a copy of this text.

77 The graphs in Fig. 9 are derived from the incidence of mortality reflected in Greek and Coptic tombstones and Greek mummy labels; see Scheidel, op. cit. (n. 38), Appendix 1, for full references. A discussion of seasonal mortality in Roman Egypt can be found ibid., ch. 1.

78 Boyaval, B., ‘Remarques à propos des indications d'âges des etiquettes de momies’, ZPE 18 (1975), 63–6;Google Scholar‘Remarques a propos des indications d'âges de l'épigraphie funéraire d'Egypte’, ZPE 21 (1976), 219, 221, 238–40;Google Scholar‘Datation du décès dans l'épigraphie funéraire de l'Egypte gréco-romaine’, Kentron 4, 3 (1988), 68–70Google Scholar.

79 G. Ferrari and M. Livi-Bacci, ‘Sulle relazione tra temperatura e mortalità nell'Italia Unita, 1861–1914’, in La popolazione italiana nell'ottocento (1985), 280 tab. 5; Shaw, B. D., ‘Seasons of death: aspects of mortality in imperial Rome’, JRS 86 (1996), 120Google Scholar fig.

80 Frier, op. cit. (n. 8), 814 tab. 6.

81 Lucr. 6.1114–15.

82 Rufus of Ephesus in Oreib., Coll. med. 44.14 (CMG 6.2.1).

83 Contis, G. and David, A. R., ‘The epidemiology of bilharzia in ancient Egypt: 5000 years of schistosomiasis’, Parasitology Today 12 (1996), 253–5;CrossRefGoogle ScholarAdamson, P. B., ‘Dracontiasis in antiquity’, Medical History 32 (1988), 204–9CrossRefGoogle ScholarPubMed.

84 S. R. Johansson, review of Benedictow, op. cit. (n. 27), Population Studies 48 (1994), 528, 531,Google Scholar also emphasized by Sallares, op. cit. (n. 41).

85 Frier, op. cit. (n. 8), 813. However, contrary to ibid., 813 n. 104, ‘a rate of that order’ is not ‘supported by the Egyptian census returns’, if only because it is impossible for a regional data set – regardless of its quality – to provide any corroboration or falsification of a general assumption about the Roman Empire as a whole. See below, n. 124.

86 Seasonal mortality patterns derived from dates of death recorded on epitaphs afford us rare glimpses of the true scale of epidemiologic and thus demographic variation in the Roman world: see W. Scheidel, Measuring Sex, Age and Death in the Roman Empire (1996), ch. 4; Shaw, op. cit. (n. 79).

87 Horden, P. and Purcell, N., The Corrupting Sea (2000)Google Scholar.

88 McNeill, W., Plagues and Peoples (1977), ch. 3.Google Scholar

89 Grmek, op. cit. (n. 34), 168–73.

90 Littman, R. J. and Littman, M. L., ‘Galen and the Antonine plague’, AJP 94 (1973), 245;Google ScholarPubMed R. Sallares, The Ecology of the Ancient Greek World (1991), 248. On the scale of the epidemic, see Duncan-Jones, op. cit. (n. 19). Thuc. 2.48 might point to a previous smallpox incursion.

91 Zelener, Y., PhD thesis, Columbia University (in preparation)Google Scholar. I am grateful to Y. Zelener for sending me a copy of his draft.

92 cf. McNeill, op. cit. (n. 88), 132, for a guess that the massive pandemic in the 250s and 260s A.D. may have been measles; cf. also Cliff, A. et al. , Measles (1993), 49Google ScholarPubMed.

93 Biraben, J.-N., Les hommes et la peste en France et dans les pays européens et méditerranéens, I (1975), 25–48Google Scholar; Conrad, L., The Plague in the Early Medieval Middle East (1981)Google Scholar.

94 Sallares, op. cit. (n. 41).

95 See above, Fig. 9, and compare Shaw, op. cit. (n. 79), 124 figs 15–16. For life expectancy in the 1920s, see The Estimation of Recent Trends in Fertility and Mortality in Egypt (1982), 12 tab. 2Google Scholar.

96 The following summary draws on Kuhnke, L., Resistance and Response to Modernization (1971)Google Scholar; D. Panzac, op. cit. (n. 44) and ‘The population of Egypt in the nineteenth century’, Asian and African Studies 21 (1987), 11–32;Google ScholarJagailloux, S., La médicalisation de l'Egypte au XIXe siècle (1986)Google Scholar.

97 Causes of death: Schnepp, op. cit. (n. 44), 552–3; Panzac, op. cit. (n. 44), 85 tab. 2. In these early datasets, only the most basic classifications (e.g., gastro-intestinal vs. pulmonary diseases) may be taken at face-value.

98 Frier, op. cit. (n. 7, 1983), esp. 331–4 for the statistical tests. The data were provided by G. Acsádi and J. Nemeskéri, History of Human Life Span and Mortality (1970), 296–7.

99 An early report assigned these burials to the period from A.D. 340 to 374: according to Frier, op. cit. (n. 7, 1983), 332 n. 9, this implies a total population of around 125 persons. However, a later study re-dated this site to the reign of Constantine I: see Lányi, V., ‘Die spätantiken Gräberfelder von Pannonien’, AArchHung 24 (1972), 138Google Scholar.

100 Frier, op. cit. (n. 7, 1983), 331. Acsádi and Nemeskéri, op. cit. (n. 98), 227, distinguish between two cranial types in this group.

101 See Acsádi and Nemeskéri, op. cit. (n. 98), 298–301 (49.5 men and 33.375 women aged 20 + ). Frier omits this telling bit of information.

102 For the problem of dual demographic ‘realities’, see above, p. 11. Frier's contribution of 1983 coincided with the onset of a wave of criticism of the feasibility of ‘paleodemography’: see esp. Bocquet-Appel, J.-P. and Masset, C., ‘Farewell to paleodemography’, Journal of Human Evolution 11 (1982), 321–33CrossRefGoogle Scholar (cf. also their ‘Paleodemography: expectancy and false hopes’, American Journal of Physical Anthropology 99 (1996), 571–83,3.0.CO;2-X>CrossRefGoogle Scholar and above, n. 52). Thus, the belief in the reliability of the demographic analysis of skeletal remains expressed by Frier, op. cit. (n. 7, 1983), 343, while common at the time, now seems rather outdated. Unfortunately, his use of paleodemographic results in op. cit. (n. 8), 790, takes no account of these intervening developments.

103 Frier, op. cit. (n. 8), 791–2 and tab. 2 (Model South Level 2 Males).

104 Hopkins, op. cit. (n. 6, 1987), 121–2 raises methodological objections to this process.

105 As I first pointed out in Scheidel, op. cit. (n. 52), 18 n. 61, the female age pattern matches Model South Level 2 Females for ages ten to twenty but increasingly deviates at later ages, especially between twenty-five and forty when implied mortality is higher even than in Level 1.

106 According to the tabulations by Szilágyi, J., ‘Die Sterblichkeit in den nordafrikanischen Provinzen’, AArchHung 17 (1965), 309–34,Google Scholar 18 (1966), 236–77, 19 (1967), 25–59, men were 1.37 times as likely as women to be commemorated in North African epitaphs (n = 17,793). For this reason alone, a significant amount of reporting bias is certain to have prevailed. Hopkins, op. cit. (n. 6, 1987), 125–6, raises the same point regarding a smaller sample from Roman Africa. Duncan-Jones, op. cit. (n. 53), 102 and n. 26, notes that age-rounding may have introduced additional distortions.

107 Hopkins, op. cit. (n. 6, 1966), 264: ‘We have to show instead some other grounds for its validity; for example, that in Africa all people who died were commemorated, hence the reasonable demographic levels of mortality at certain ages. But clearly we do not have enough evidence for such an assertion.’ This standard has not been met: cf. Frier, op. cit. (n. 8), 791: ‘In the case of the European inscriptions, no life table based on all or part of them is even remotely plausible. (…) Roman North Africa is altogether different. (…) they produce credible mortality rates for males aged 10 to 44, and for females aged 10 to 54’, etc., a line of reasoning that illustrates Hopkins' principle. This approach also undermines the view that ‘convergence of the best available statistics still demands a measure of respect’ (Bagnall and Frier, op. cit. (n. 7), 109), given that there is no quality standard beyond goodness of fit with Coale/Demeny models. Cf. furthermore Scheidel, op. cit. (n. 52), 20–1.

108 I am disinclined to discuss under this heading the evidence of Ulpian's so-called ‘life table’ (Dig. 35.2.68 pr.) (exploited by Frier, op. cit. (n. 7, 1982)) since it remains unclear whether this text is based on any kind of empirical evidence at all (pace Duncan-Jones, op. cit. (n. 53), 100–1). For scepticism, cf. Parkin, op. cit. (n. 6), 27–41, 75–8, 82–3; Hopkins, op. cit. (n. 6, 1987), 120–1; Saller op. cit. (n. 1), 13–15.

109 Bagnall and Frier, op. cit. (n. 7), 84–90. For the pitfalls inherent in this extrapolation process, see above, Section II.

110 I hasten to qualify this statement with a reference to the Chinese data cited in n. 49, which show marked gender differences.

111 For city and village population numbers, see Rathbone, D., ‘Villages, land and population in Graeco-Roman Egypt’, PCPS 36 (1990), 103–42Google Scholar.

112 The latter deviation is statistically significant. In the cities, 71.1 per cent of residents aged five and over are between five and thirty-four years old, compared to 64.7 per cent in the model (p < 0.0208). 62 per cent of those older than fifteen are aged between fifteen and thirty-four, compared to 53 per cent in the model (p < 0.0066).

113 In general, see Braunert, H., Die Binnenwanderung (1964)Google Scholar.

114 Muniruzzaman, A. N. M., Demographic Survey in East Pakistan, 1961–1962, part 3 (1966), 9–10Google Scholar.

115 See Scheidel, op. cit. (n. 38), ch. 1.3.2 for discussion.

116 See above, n. 69. Bagnall, R. S., Egypt in Late Antiquity (1993)Google Scholar, 50, likewise assumes that cities were less healthy than villages. This possibility is not investigated in Bagnall and Frier, op. cit. (n. 7).

117 Temporary migration would tend to be sex-specific. In the absence of obvious concealment of village women, there is no reason to believe that the ranks of young adult women in the cities were boosted by temporary migrants. In the villages, women at these ages consistently outnumber men: Bagnall and Frier, op. cit. (n. 7), 163 fig. 8.2.

118 However, this is the implicit assumption logically underlying the weighting procedure in Bagnall and Frier, op. cit. (n. 7), 82, where they recalculate figures on the assumption that the ratio of urban to rural population was 1:2 (i.e., giving village data twice the weight of urban data), in the apparent belief that this would restore a representative sample of the actual population.

119 If the sample of 141.1 female villagers (based on seven-year moving averages) aged fifteen to sixty-nine is divided into two roughly equal halves, 51.45 percent are younger than thirty-five. In terms of statistical significance, this proportion is compatible with corresponding shares predicted by Model West Females Levels 1 (p < 0.4) through 6 (p < 0.516).

120 See above, n. 112.

121 See above, nn. 69–70.

122 It should be noted that permanent immigration at the onset of maturity would have had no palpable impact on the urban distributions in Figs 12–13, except for the possibility that recent immigrants were particularly vulnerable to unfamiliar urban disease environments.

123 Frier, op. cit. (n. 8), 789.

124 e.g., Frier, op. cit. (n. 7, 1983), and Bagnall and Frier, op. cit. (n. 7), esp. 84–90. Attempts to derive typical rates of population growth from small data samples are particularly hard to defend. Natural growth rates need to be known from independent sources (such as successive population counts) before attested age distributions can be converted into life tables. It is technically impossible to derive both age-specific mortality rates and growth rates from the same body of data: for instance, high mortality and strong growth will shape age structure in the same manner (by increasing the proportion of young people). This insuperable problem invalidates the claim that the (entirely plausible) estimate of an average long-term annual growth rate of 0.15 per cent in the early Roman Empire ‘is supported by the Egyptian census returns’ (Frier, op. cit. (n. 8), 813 n. 104), a claim based on the interpretation of subtle differences between the adult age distribution of women in these documents and one particular model life table (Bagnall and Frier, op. cit. (n. 7), 86). This claim is predicated on the notions that (1) the female age data in the census texts can be shown to be representative of the Egyptian population (which is not the case: see above, Section V); (2) the chosen model life table provides a normative standard (which is unlikely: see above, Section II); (3) local growth rates coincide with global averages (which is even more unlikely: see above, Section IV). The assertion that ‘the age-distribution at Keszthely-Dobogó is generally consistent with a population declining at a rate of somewhat less than.5% per year' (Frier, op. cit. (n. 7, 1983), 331 n. 5) is also voided by this bundle of problems. In fairness, I ought to point out that some of my own earlier work is vitiated by an overly optimistic application of model life tables to inadequate data samples (see esp. Scheidel, op. cit. (n. 86), 117–24).

125 Frier, op. cit. (n. 8), 791.

126 Outright scepticism about the relevance of model life tables has been rare among ancient historians: see, however, M. Golden, ‘A decade of demography: recent trends in the study of Greek and Roman populations’, in P. Flensted-Nielsen et al. (eds), Polis and Politics (2000), 32; Sallares, op. cit. (n. 41).

127 Hopkins, op. cit. (n. 6, 1966), 264.

128 According to the Human Development Report 1999 (1999), 137Google Scholar, e0 in war-torn Sierra Leone was 37.2 years in 1997, the only country with a reported rate of under 40. By now, the worst AIDS-affected countries such as Zambia are about to drop below 40 years.

129 Based on Model West Levels 1 and 8/9 Males, respectively. Of course, this calculation depends on the age structure predicted in these models.

130 As a consequence, the number of orphans in South Africa may rise to 2 million by 2010 (The Economist, 24 February–2 March 2001); this implies that the proportion of minors without living parents will be roughly twice as large as in Saller's Roman kinship simulation based on Model West Level 3 Females (Saller, op. cit. (n. 1), 49), even though mean life expectancy at birth in South Africa will be forty years instead of twenty-five as in the model.

131 Golden, op. cit. (n. 126), 32.

132 Saller, op. cit. (n. I), ch. 3.