Tim G. Parkin, Demography and Roman Society. Baltimore and London: The Johns Hopkins University Press, 1992. Pp. xvi, 225. $29.95. ISBN 0-8018-4377-4.
Reviewed by Bruce W. Frier, Institute for Advanced Study.
The Multilingual Demographic Dictionary defines demography as "the scientific study of human populations, primarily with respect to their size, their structure and their development." 1 Demography is "scientific" in the sense that demographers seek to understand a population from an objective standpoint, above all through statistics that describe it. Statistical standards are naturally rigorous for modern populations of developed nations, but more relaxed for less developed and historical populations, where information is often incomplete and unreliable. But whatever their form and quality, statistics are the lifeblood of demography, and nothing can substitute.
Tim Parkin's book, which evolved from his Oxford dissertation, has two considerable merits. The first is that he forthrightly confronts the statistical problem as it applies to the Roman Empire. Surviving sources rarely provide statistics at all, much less what we would think of as entirely trustworthy statistics. In his opening chapter, Parkin is at pains to show that all available ancient sources -- the epitaphs giving age at death, the Egyptian census returns, Ulpian's life table, the small corpus of analyzed skeletons -- are uniformly "so plagued with biases and produce such potentially misleading or improbable information that they cannot be considered as usable" (p. 58).2 As we shall see, this conclusion is hyperbolic. But Parkin's self-described "caution and skepticism (some would call it pessimism)" (p. 134) are nonetheless often salubrious even when he covers thoroughly familiar ground, as with the epitaphs (pp. 5-19).
The book's second merit is that Parkin has a secure grasp on the techniques of modern demography, especially the modelling methods that demographers now use when discussing incomplete or suspect statistics. Parkin's second chapter presents a clear introduction to the subject. Presuming no knowledge of mathematics, he explains how basic demographic functions of mortality and fertility are derived, as well as how they can be used. Non-numerate classicists will find his account "user-friendly"; and the bibliography is also full and accurate.3 Parkin concentrates on mortality and fertility, which are jointly the major determinants of age and sex structure. By contrast, he has little to say about the third major demographic function, migration (pp. 135-136), and next to nothing about gross population statistics, household structure, and other matters often included in demographic studies. But what he does discuss goes to the heart of the matter.
The third chapter, ominously titled "Demographic Impressions of the Roman World," offers heterogeneous comments on aspects of Roman mortality and fertility. Although Parkin breaks no new ground here, he does show how a knowledge of demography can help ancient historians in interpreting literary and other evidence. Indeed, throughout the book he offers generally courteous but often devastating criticism of the casual manner in which many ancient historians have handled demographic issues.4 Scholars both young and old can learn from his discussion. In fact, there is now no excuse for not learning.
Unfortunately, when all is said and done, Parkin leaves the central demographic issues just as murky as ever. The reason, of course, is his firm disavowal of all ancient statistical evidence, a disavowal that renders him helpless when attempting to reconstruct the Roman demographic regime. Eschewing ancient evidence, Parkin relies instead on comparative evidence in order to choose an illustrative model for Roman mortality (pp. 84-86, 92). He argues that average Roman life expectancy at birth is likely to have lain between 20 and 30 years,5 and he then opts for a value "midway" between these two extremes. Splitting the difference may seem fair enough, but in truth this is nothing but idle speculation. Although, from a modern perspective, all these values may look alike, a life expectancy of 30 years is fifty percent higher than one of 20 years, an enormous difference with large consequences especially for fertility. Demography simply cannot proceed by just knocking models around, without at least a moderately firm statistical foundation.
As Parkin realizes, I have been a leading proponent of the view that, despite the obstacles, some ancient evidence can be used to establish at least the most likely mortality functions.6 Inevitably, my view comes in for a good deal of fair but trenchant criticism (pp. 27-41, 82-85). I am unrepentant. There is nothing to repent. In the last analysis, Parkin's obstinate skepticism,7 though it doubtless gratified his Oxford mentors, is simply too coarse to signify.
The best available ancient evidence indicates that even 25 years is too high as an estimate for Roman life expectancy at birth. This conclusion is bolstered not only by Ulpian's life table,8 but also by at least a large section of the mortality statistics (ages 5 to 50) derived from African epitaphs;9 the carefully analyzed Hungarian cemeteries may provide some additional support (pp. 41-58). And further evidence has since emerged: the age distribution in the Egyptian census returns implies a female life expectancy at birth of approximately 22.5 year, in a stable population growing at an intrinsic rate of about 0.2 percent per year.10 Despite cavils, even Parkin recognizes that the census returns are by far the most credible surviving demographic evidence from antiquity (pp. 19-22), though he makes virtually no use of them in his book.11
What matters here, I should stress, is not the precision of any individual piece of evidence, but rather the convergence and cumulative weight of the evidence, a point that Parkin entirely misses. Seriatim demolition of our major sources just does not advance the argument, which would, in fact, have been better served if Parkin had reversed the order of his first two chapters. Once modern demographic methods are in place, the ancient evidence, despite its idiosyncrasies and inadequacies, begins to reveal latent but highly eloquent patterns. By upending the logical order of his discussion, Parkin shifts the burden of proof, illegitimately but with clear calculation (pp. xii-xiii); this is a debater's skill, not a scholar's.
Still, why would anyone quibble about a few years of life expectancy, one way or the other? There are two main reasons: first, life expectancy at birth is the single most widely used measure of a population's general social welfare; second, if in truth Roman life expectancy lay in the lower 20's, then it was below even the lowest attested early modern European levels. This would mean that, by one significant measure, the Roman Empire achieved little if any social benefit for its subjects -- something that, if true, seems well worth the knowing.12
At one point in his book, Parkin quotes with approval Oldenziel's view: "It is dubious whether demographic arguments are any more precise or conclusive than arguments based on literary sources."13 Oldenziel is right. But her point actually cuts against Parkin. Ancient historians work every day with sources that are biased or misleading; so also do historical demographers. The art of both disciplines is to diagnose causes of bias and (if possible) to correct for them; it is not to weed out sources deemed "defective" in some respect or another. After all, nobody seriously suggests eliminating Tacitus as a source for the reign of Tiberius; but what Parkin proposes for Ulpian's life table or the Egyptian census returns is every bit as radical. At any rate, I can't see the difference, unless it has something to do with the "innate fear of figures and equations" that Parkin mentions in his preface (p. xiv). But I dare say this fear is a product more of nurture than of nature.
Parkin is fundamentally (though I hope not irretrievably) confused about the distinction between two questions: can we derive demographic functions from ancient statistical sources, and, if we can, how much confidence should we then place in the results? The answer to the first question is, or ought to be, much more positive than the answer to the second. In any event, unlike Parkin, I would sharply distinguish caution from skepticism. A cautious scholar looks both ways before crossing the street. A skeptical scholar doubts that there is a street. Judge for yourself which is more likely to be hit by a bus.
 E. van de Walle, Multilingual Demographic Dictionary, English Section (1982) 101.  This view broadly echoes K. Hopkins, "Graveyards for Historians," in La Mort, les Morts, et l'Au-dela dans le Monde Romain (ed. F. Hinard; 1987) 113-126.  The next step up from Parkin is C. Newell, Methods and Models in Demography (1988), an excellent introduction.  E.g., pp. 22-27 on A.E. Samuel et al., Death and Taxes: Ostraka in the Royal Ontario Museum (1971). At times Parkin is outright patronizing; e.g., p. 74 on Brent Shaw.  This range is too narrow; pre-modern populations not infrequently attained life expectancy up to 40 years, see M. Livi-Bacci, Population and Nutrition: An Essay on European Demographic History (trans. T. Croft-Murray, 1991) 1-10.  B.W. Frier, "Roman Life Expectancy: Ulpian's Evidence," HSCPh 86 (1982) 213-251, and "Roman Life Expectancy: The Pannonian Evidence," Phoenix 37 (1983) 328-344. See also my article on "The Demography of the Roman Empire," to be published (conceivably within my lifetime) in Cambridge Ancient History vol. XI; the article was written ten years ago.  E.g., p. 59: "The fact of the matter is that there is a general lack of vital statistics from the ancient world," with a note citing M.I. Finley and A.H.M Jones. I need hardly say that this view is unsound; see recently, e.g., R.P. Duncan-Jones, Structure and Scale in the Roman Economy (1990).  See note 6. Ulpian's life table is in Aemilius Macer (2 ad Leg. Vic. Hered.), D. 35.2.68 pr. Parkin believes these figures are "based on good guesswork" (p. 38) or "intelligent guesswork" (p. 40); why that should not suffice to give them some credibility, Parkin declines to say. R. Duncan-Jones, (cited n. 7) 101, agrees with me that Ulpian's life table gives "relatively plausible" if "very crude" values, at least for younger ages.  As Parkin concedes, pp. 165-166, n.51. The African epitaphs are of interest because, unlike in Roman Europe, age of death is almost invariably included; this largely eliminates a major source of statistical bias.  These estimates come from R.S. Bagnall and B.W. Frier, The Demography of Roman Egypt, work in progress.¬  Likewise for Egyptian tax lists with ages of adult male tax-payers; Parkin knows only one such list (pp. 21-22), which he peremptorilly dismisses because "the sample remains too small to be truly meaningful." In fact, there are five tax lists, with about 600 ages; these lists fully support the census returns.  G. Acsádi and J. Nemeskéri, History of Human Life Span and Mortality (1970) 216: Roman "mortality characteristics do not differ substantially from those of the Eneolithic or Bronze Age."  Parkin, p. 68, quoting R. Oldenziel, "The Historiography of Infanticide in Antiquity: A Literature Stillborn," in Sexual Asymmetry: Studies in Ancient Society (ed. J. Blok and P. Mason; 1987) 87-107, at 98.