1
Proxies for Economic Performance in the Roman Empire
KEITH HOPKINS in a seminal article, “Taxes and Trade in the Roman Empire (200 BC–AD 400),” presented a structural argument for modest growth in the Roman economy, corroborated by proxies implying an increase in “general prosperity.” His proxies—the chronological distribution of shipwrecks and the quantity of coinage—were substitutes for the lack of aggregate statistics for the Roman imperial economy. Over the past four decades these proxies and others have been proposed and critiqued, and despite the criticisms the search for viable proxies to measure the success of the Roman economy has continued. This chapter will summarize the past debate and point to weaknesses in some of the very recent proposals for proxies in support of the more general argument of this book that none of the proxies is reliable enough to justify neglect of literary sources such as Pliny’s Natural History.1
Why have Roman economic historians persisted in trying to identify proxies as indices of aggregate performance despite the criticisms that repeatedly followed? The motives of different historians no doubt vary, but at the heart of the matter is the fundamental question of whether the Roman Empire was benign or oppressive and why it ultimately failed. Was Gibbon right in his famous assertion that “if a man were asked to fix the period in the history of the world, during which the condition of the human race was most happy and prosperous, he would, without hesitation,” point to the years 96 CE to 180 CE.2 Roman conquests were undoubtedly brutal, but did Roman rule also, in the long run, make inhabitants of the empire “safer” and “richer”?3 And if so, how much safer and richer? The level of prosperity of the empire has a bearing on our evaluation of the approaches of New Institutional Economics. Can we demonstrate with proxies that Roman peace, infrastructure, and law had the net effect of continually improving productivity and enriching the population? Or on the other hand, can proxies demonstrate that oppressive Roman institutions eventually put a drag on the economy, as some prominent modern economists have claimed?4 The current optimistic view is that the Pax Romana nurtured sustained growth until the external shock of the Antonine Plague (165 CE) sent that growth trend into reverse. An alternative view is that in the absence of substantial technical breakthroughs to increase per capita productivity the empire could not achieve sustained growth for endogenous reasons. The conclusion of this chapter is that despite the increasing sophistication of the arguments for proxies for growth, the most recent efforts do not hold up to close scrutiny and so do not demonstrate sustained growth up to the Antonine Plague.
Hopkins’s original article, later revised and updated, presented graphs of dated shipwrecks and minted coins in circulation. The graph of shipwrecks peaked sharply in the last two centuries before the common era, declined somewhat in the next two centuries, and then a steep decline set in. The graph of silver coins in circulation, like that of shipwrecks, displayed a steep rise in the second century BCE.5 Hopkins was acutely aware of the limitations of the evidence and his interpretation, acknowledging that “all these arguments, and the evidence from which they are derived, are partial, but they draw strength from their inter-relationship.”6
In the 1990s a set of articles by Hong and colleagues adduced data from environmental science to demonstrate a rise in mining activity. Their measurements of lead pollution frozen in the Greenland ice core showed that anthropogenic lead pollution rose well above natural background levels during the classical period, reaching a peak around the turn of the era. The pollution, it was argued, came primarily from silver mining in Spain in which lead was vaporized in the process of cupellation to refine the ore. The conclusions were based on fourteen sections from the ice core distributed across approximately six hundred years of the Roman Republic and Empire.7 Hopkins later incorporated these findings as an additional proxy for economic performance: “These scientific estimates of ancient pollution and total production give us an unprecedented vision of economic growth and inefficiency in classical antiquity.”8 The data series from Hong and colleagues with fourteen measurements was not sufficiently fine-grained to provide more precise dates for the decline of mining and smelting. Refinement of the data was to come two decades later (see p. 15).
In a much-cited article of 2005, François de Callataÿ summarized the argument from the proxies in the following conclusion:
The convergence of these graphs [showing atmospheric lead pollution, atmospheric copper pollution, and shipwrecks] clarifies an increased activity during the last centuries B.C. and the first three centuries A.D. Of the three, the wrecks is the most spectacular with its peak in the period 200 B.C.–A.D. 200.… Within the margin of a century it is also the one spoken of by J. Andreau when he defined the period of greater prosperity of economic life as being that which runs “du IIe siècle av. J.-C. au premier siècle de l’Empire” [from the second century BCE to the first century of the empire]. While lacking the clarity and precision of that of the wrecks, the charts for lead and copper move in the same direction: they are characterized by significant variations from one measurement to another, which leads us to hope that one day we might possess several complete sets of data spread out over nearly three millennia. For the present, it is wise to resist the temptation to tie such-and-such a high value obtained for such-and-such a year with any known or supposed historical events.9
De Callataÿ’s caution was wise, and his hope for more refined data was soon to be satisfied.
About the same time, Willem Jongman and others pointed to another proxy, based on research by Nobelist Robert Fogel and Dora Costa, who identified stature as an index of human health and well-being.10 In recent centuries populations have grown taller generation by generation as they have grown wealthier, the duration of life has been extended, diets have improved, and many diseases have been brought under control. Jongman in 2007 concluded that “Gibbon was right” based on a study of femur bones excavated by archaeologists: “the rise [in the length of femur bones] is spectacular—the first and early second century peak equates European stature in the early twentieth century. Decline clearly set in in the late second century A.D., to be followed by a recovery later in the third century, and ultimate collapse with the fall of the Western Empire.”11 Whereas the previous studies did not try to tie the proxies to a political history narrative, Jongman’s reference to Gibbon placed the turning point in general prosperity in the late Antonine era with consequences for explanations of exogenous as opposed to endogenous causes. A decade later Jongman reversed his conclusion about the trends in stature and health and offered a different explanation (see p. 19–21).
In 2009, Walter Scheidel and Andrew Wilson engaged in a debate that sharpened the focus on the issues and articulated detailed criticisms of previously suggested proxies. Whereas Hopkins and de Callataÿ had deliberately assigned the timeline of growth and decline in terms too vague to distinguish exogenous from endogenous causes, Scheidel drew attention to the consequences for quite different explanations: “In the most basic terms, we must ask whether Roman economic growth was terminated by endogenous or exogenous factors. Was economic development the result of a limited and unrepeatable improvement in productivity that was gradually offset by population growth …, or was productivity growth theoretically sustainable in the long run but in practice curtailed by shocks to the system such as political or military crises?”12 Of course, the other possible external shock was the Antonine Plague. Scheidel highlighted the weaknesses of each of the proxies as a barometer of productivity growth: the trend in number of shipwrecks was influenced by multiple biases; lead pollution as a measure of silver production need not be related to economic performance as shown by the comparison with early modern Spain; trends in animal bones could reflect changes in taste rather than levels of meat consumption. For these and other reasons,13 the historian should be wary of taking any of these proxies as a reflection of economic performance, but if one does take them seriously, then the appropriate method is to analyze rate of change, not simply to identify the peaks: “Although overall levels of economic performance are by no means irrelevant, from an economic perspective the latter question [rate of change] is of even greater importance.” And “in so far as it is legitimate to put much weight on these observations …, they would seem to suggest that economic expansion stalled sometime around the turn of the era,” not the later second century in the aftermath of the plague.14 That is, the enrichment was a product of conquest and the consolidation of the empire and was not sustainable.
Scheidel then presented a model to show that, based on various plausible assumptions about population growth and imperial expenditures, even a modest, sustained growth rate over a period as long as the Pax Romana makes it hard to understand the strain on imperial finances in the late second and third centuries: “this matrix makes it hard to posit intensive growth of closer to 0.2% than to 0.1% (or indeed even 0%) per year, let alone a higher rate.”15 The article acknowledges increases in the richness of the archaeological record during the centuries of Roman rule but again suggests caution about inferring growth in productivity based on the comparison with early modern Europe when increasing inequality, rather than increasing productivity, produced a lavish material heritage.16 With due circumspection, Scheidel concluded that the hypothesis of a one-off increase in prosperity resulting from the consolidation of Roman rule was more likely than one of sustained growth abruptly halted by the Antonine Plague.
Wilson’s careful response to Scheidel, based on his archaeological expertise, served to underline the uncertainties about several of the proxies (shipwrecks, animal-bone data, and lead pollution).17 As a result, Wilson suggested that even Scheidel’s conditional hypothesis based on taking the proxies seriously should be rejected and that “we are still a long way off the construction of sufficient robust proxies to generate an argument from convergence, though we are making progress.”18 He then proposed other proxies: dated building inscriptions and honorific inscriptions, on the grounds that “the overall spending pattern [on public buildings] may well bear a meaningful relationship to overall economic performance.”19 It seems to me that both of these indicators are vulnerable to the same doubts as others—that is, they were subject to cultural influences as well as economic capacity. In his important article “The Epigraphic Habit in the Roman Empire,” Ramsay MacMullen pointed to the “cultural significance” of erecting inscriptions and warned against the assumption “that the body of all inscriptions against which attestation is measured does not itself rise or fall—a false assumption. So administrative, economic, social, and religious history needs to be rewritten.”20 Apart from changing cultural preferences, increasing inequality of wealth could have influenced the trend.21 In addition, the claim that “large sums” were spent on the public buildings needs to be put in context: J. W. Hanson’s inventory registers an increase in urban monuments from 1,803 to 3,928 over the four centuries from 100 BCE to 300 CE, or an average of about five per year; R. Duncan-Jones’s price lists put the median construction cost of buildings in Italy at 300,000 HS and 43,500 HS in the African provinces—a small fraction of total production, even if the number of monuments and the costs are greatly understated.22 In his 2009 article, Wilson concluded that it is immensely difficult to formulate models that allow an unambiguous choice between hypothese...