Hostname: page-component-586b7cd67f-2brh9 Total loading time: 0 Render date: 2024-11-23T19:33:56.777Z Has data issue: false hasContentIssue false
  • English
  • Français

Taxes and Trade in the Roman Empire (200 B.C.–A.D. 400)*

Published online by Cambridge University Press:  24 September 2012

Keith Hopkins
Keith Hopkins
Affiliation:
Brunel University
Get access Rights & Permissions [Opens in a new window]

Extract

This essay is speculative and tentative, a preliminary attempt at exploring a broad territory of Roman economic history over a long period. For the sake of clarity, I have canvassed several probabilities in the form of propositions, but the evidence is so sparse that it is difficult to prove that each proposition is right. It is disappointing to confess at the outset that one's case is unproven and that the generalizations advanced are disproportionately large in relation to the supporting evidence. Even so, the experiments made here with both evidence and methods may stimulate others into refuting or reshaping the propositions. And besides, some of the methods can be usefully applied to other problems in Roman history.

Type
Research Article
Information
The Journal of Roman Studies , Volume 70 , November 1980 , pp. 101 - 125
Copyright
Copyright © Keith Hopkins 1980. 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.)

References

1 I shall concentrate here on argument and on the economic structure of the Roman empire, rather than on what Romans thought they were doing or on surviving sources. I have adopted this tactic simply because I want to cover a broad canvas in a relatively short article, not because I feel that the Romans' own economic thoughts or writings should be neglected. But it does mean that some of the conventional sign-posting is missing.

2 I must stress the correlative form of the second proposition: In so far as …, then … (in so far as …). For the moment, I make no estimate of the volume of taxes, nor of the extent to which they were raised in money.

3 These two propositions also imply an explanation for the increase of imports into Italy during the High Empire. I do not mean that Italy stopped exporting, only that the balance of trade favoured imports; the explanation is to be sought more in economic forces than in an Italian moral decline.

4 The problem of how much staple food was transported, long-distance overland or by ship, cannot be solved simply by pointing out the high relative cost of land transport. That alone did not make it absolutely prohibitive. I suspect that availability of transport, information and trader organization were also important. Comparative evidence illustrates the problems. In Italy in the sixteenth century, staples were occasionally transported from the eastern coast of Italy overland to the city of Rome (Delumeau, J., La vie économique et sociale de Rome (1959) 11, 521 ff.Google Scholar), but in southern France in the same period, the volume of transport available for carting supplies between towns was too small to even out inter-city variations in price (Baehrel, R., Une croissance: la Basse Provence rurale (1961), 531 ff.Google Scholar).

5 Archaeological evidence is uncorrectably biassed by the survival of pots, which cannot have been so important in the Roman economy. The salvation is that surviving pots can reasonably serve as proxy for perishable goods such as textiles, which have not survived archaeologically but which probably were important economically. Thus distribution maps of pots illustrate the viability and direction of long-distance trade. For one example see Peacock, D. P. S. in du Plat Taylor, J. and Cleere, H. (edd.), Roman Shipping and Trade: Britain and the Rhine Provinces, The Council for British Archaeology Research Report 24 (1978), 49Google Scholar.

6 On the growth of towns, see especially Finley, M. I., ‘The Ancient City: From Fustel de Coulanges to Max Weber and Beyond’, Comparative Studies in Society and History 19 (1977), 305 ff.CrossRefGoogle Scholar; Hopkins, K., ‘Economic Growth in Towns in Classical Antiquity’, in Abrams, P. et al. (edd.), Towns in Societies (1978), 35 ff.Google Scholar, where several of the issues discussed here are put in a different form.

7 Once again, I am for the moment concerned more with the logic of the argument than with the evidence by which one could allocate regions conquered by the Romans along a continuum of economic sophistication, or lack of it. I am certainly not assuming that all western provinces were economically primitive before the Roman conquest. I suspect (though how would one prove ?) that they did become more sophisticated after conquest by the Romans. Cf. note 13 below.

8 The concept, general prosperity, is purposely vague. We know that the rich were rich, and we admire great public buildings, such as the theatre at Aspendos or the temples at Petra and Palmyra. But how can we know about the distribution of wealth, and the standard of living of relatively poor townsmen or peasants?

9 ‘Traders roam from sea to sea looking for some market which is badly stocked’, so Philostratus, , Life of Apollonius of Tyana 4. 32Google Scholar. Inter-annual fluctuations of rainfall have been largely ignored by ancient economic historians, perhaps because no ancient source mentions them. See the International Year-book of Agricultural Statistics for modern national figures.

10 The most exciting documents which illustrate the nature of ancient Mediterranean trade are the Geniza papyri from Cairo, dating from the tenth century A.D. onwards (see Goitein, S. D., A Mediterranean Society 1 (1967))Google Scholar. Nothing from the classical period can rival them. In spite of their late date, they are useful for Roman historians. The long lists of produce in T. Frank, ESAR, passim, reflect an antiquarian idea of what economic history should be. The main questions should be: which organizaions of traders, by what mechanisms (partnership, investment, credit, cash?), sold how much of what to whom? Even the customs lists at Zarai, Numidia (CIL VIII, 4508) and Palmyra (OGIS 629) do not help us reach an answer to these questions.

11 Augustus, Res Gestae 15; D. van Berchem, Les distributions de blé et d'argent à la plèbe romaine dans l'empire (1939); cf. the Chinese experience in supplying Peking: Hinton, H. C., The Grain Tribute System of China (1956)Google Scholar.

12 Two cautions. First, not all peasants paid rent, and die categories rentier/free holder/tenant overlap, since many small landowners in the course of the family cycle supplemented their livelihood by renting out surplus land or by renting it in. This is clear from the evidence of Roman Egypt and is explained theoretically by Chaianov, A. V., The Theory of Peasant Economy (1966)Google Scholar. Secondly, I do not wish to imply that the surplus was fixed in size. Indeed, I argue that the demand for taxes and rent probably increased the size of the surplus produced. Moreover, the concept ‘disposable surplus’ is an objective account of what was produced over and above minimum subsistence. Peasants themselves may not have thought of it as surplus, although the concept did exist in classical times.

13 This can be illustrated, but not, I think, proved. Roman levels in excavations reveal more artefacts than pre-Roman levels: more coins, pots, lamps, tools, carved stones and ornaments—in sum, a higher standard of living. Since archaeologists seem very reluctant to write synoptic works, I cite four corroborative illustrations from different regions: Clavel, M., Bezières (1970), 332Google Scholar; Schulten, A., Geschichte von Numantia (1933), 154–5Google Scholar: Callu, J.-P., Thamusida (1965), 187 ff.Google Scholar; Kraeling, C. H., Ptolemais (1962), 270 ff.Google Scholar

14 Were rents paid in money? Wealthy landowners living in the city of Rome clearly needed large amounts of money to spend, as well as produce from nearby estates. Cicero (Paradoxa Stoicorum 49) expressed income in money terms, not in wheat, and much later Olympiodorus (frag. 44) declared that Roman aristocrats in the fourth century A.D. received one quarter of their incomes in kind. Income from large estates given to the Roman church by Constantine (Liber Pontificates, ed. L. Duchesne (1886) 1, 170 ff.) was also mostly in money, and some of what was to be paid in kind was not grown on the estates but had to be bought in the market. The Igel monument of the third century A.D. does not show money payments by tenants, but payments made to workers—see the convincing arguments by Drinkwater, J. F., Trierer Zeitschrift 40/1 (1977/1978), 116Google Scholar. My fragmentary illustration reflects the neglect of rent by Roman historians.

15 Parker, A. J., ‘Ancient Shipwrecks in the Mediterranean and the Roman Provinces’, British Archaeological Reports, Supplementary Series (forthcoming) (1980)Google Scholar; cf. Progress in Underwater Technology 4 (1979)Google Scholar. I am most grateful to Dr. Parker for letting me know about his important findings before their publication.

16 See now Pomey, P. and Tchernia, A., ‘Le tonnage maximum des navires de commerce romains’, Archaeonautica 2 (1978), 233 ff.CrossRefGoogle Scholar, who show that there is now sufficient archaeological and legal evidence (D. 50. 5. 3) to indicate that ships of about 400 tonnes burden and over were commonly used from the last century B.C., a tonnage not reached again until the fifteenth century by Genoa and Venice. I am assuming that average size was influenced by the construction of these large ships, but no direct evidence on average size exists, pace L. Casson, Ships and Seamanship in the Ancient World (1971), 183 ff., who relies too heavily on IG XII, Suppl. 348, which has been doubtfully restored in the critical passage—cf. BCH 57 (1933), 394 ff.

17 See Parker, op. cit. (n. 15). The unfortunately long time-periods used in Figure 1 reflect the crudity of dating available. Still, more intervals would create more problems at the boundaries. I imagine there may have been significant variations within each long period. It has been suggested to me that more ships sank in the central period because of the Roman penchant for transporting heavy loads—marble, amphorae—by ship. It may be so, but this argument also illustrates scholarly ingenuity when confronted with a plausible generalization.

18 Crawford, M. H., Roman Republican Coinage (1974), 694Google Scholar. The evidence supporting the conclusion that 30,000 coins were struck per die, normally, is impressionistic and plausible, but by no means certain. See contra, Mattingly, H. B., ‘Coinage and the Roman State’, Numismatic Chronicle 17 (1977), 206 ff.Google Scholar, arguing for 15,000 denarii struck per obverse die and for lower military costs. In my judgement, Crawford wins the argument, on points (but see n. 29).

19 Why would the Roman mint systematically take in partly worn old coins, of almost pure silver, and remint to heavier new coins of the same purity? By doing this, the mint would shoulder all the cost of wear and of reminting. The answer depends partly on the fiduciary element in the currency, on how far coins were valued above their silver content, and on the availability of silver bullion. According to Polybius (Strabo 3. 2. 10), the Roman state in the mid-second century B.C. received 25,000 drachmae per day from the silver mines at New Carthage, Spain, in which 40,000 men worked. This comes to c. 35 tonnes per year for 365 days.

20 Patterson, C. C., ‘Silver Stocks and Losses in Ancient and Modern Times’, Economic History Review 25 (1972), 207–10CrossRefGoogle Scholar.

21 If we increase the rate of loss to 10 per cent per year, the overall rise in silver coin stock 157–77 B.C. is still fivefold; the rise during the second century becomes slower, but the fall in the money supply from 77 to 50 B.C. becomes dramatic—more than 50 per cent. Surely, it is too dramatic to be credible. So is the implied absolute loss of coins.

22 Both the high and the low estimates were probably on the high side. The high estimate was based on the following assumptions:

200–158 B.C.: the equivalent of one million denarii p.a.

220–201 B.C.: the equivalent of two million denarii p.a.

240–221 B.C.: the equivalent of one million denarii p.a.

The low estimate was half these levels. A loss rate of 2 per cent per year was also assumed. The result was 46 million and 23 million denarii respectively for the stock of silver coins in circulation in 158 B.C. The guesswork in these crude calculations hardly needs stressing. Please note that Crawford had already underlined the low volume of silver coins minted in the decade before 157 B.C., op. cit. (n. 18), 625.

23 If, for example, we began with a stock of 80 million denarii in 158 B.C. instead of 35 million denarii, again with a loss rate of 2 per cent per year, then the total silver coin stock still rises sixfold to its peak in 77 B.C., although the rate of growth in the second century B.C. is slower. The general trend remains similar.

24 Crawford, M. H., Roman Republican Coin Hoards (1969)Google Scholar.

25 idem, ‘The Financial Organization of Republican Spain’, Numismatic Chronicle 9 (1969), 84.

26 Compare, for example, the fifteen British coins found at Maiden Castle with the several hundred Roman coins found at Verulamium (Wheeler, R. E. M., Maiden Castle (1943), 329Google Scholar; Verulamium (1936), 223 ff.). But are the sites comparable? In any such comparison there are some difficulties. And besides there were exceptions: some districts of Britain had come under Roman commercial influence before the conquest. Even so, in spite of the difficulties of illustrating it, the generalization still holds, I think.

27 See G. Parker in Cipolla, C. M. (ed.), Economic History of Europe (1974) 11, 527–8Google Scholar; on the huge scale of Roman silver mining, see Blazquez, J. M., ‘Explotaciones mineras en Hispania durante la República y el alto Imperio Romano’, Anuario de Historia económica y social 2 (1969), 9–68Google Scholar; Avery, D., Not on Queen Victoria's Birthday (1974), 413 ff.Google Scholar; also Patterson, op. cit. (n. 20), 218 ff.; and Hopkins, op. cit. (n. 6), 56.

28 Ardant, G., Histoire de l'impôt (1071) I, 114Google Scholar; cf. Théorie sociologique de l'impôt (1965).

29 Crawford, op. cit. (n. 18), 694, cf. 617 and 633. I cannot agree with Crawford's suggestion that soldiers were typically paid with new coin, even in the second century B.C., let alone that minting purposively matched state expenditure on the army. Many soldiers served too far away from the city of Rome to be supplied from there with new coin, and besides the annual mintage of new coins constituted only a small proportion of all the coins in circulation. Why pay in new coins only ?

30 To be fair, Crawford did not strictly make a proposition; he just drew a conclusion, and expressed it sufficiently clearly so that it could be tested. If he had proposed a general relationship between military expenditure and minting which took time to show, then we could have done a lagged correlation: military expenditure in years 1, 2, 3 with minting in years 2, 3, 4, etc. But that is not what he suggested.

31 M. H. Crawford, ‘Coin Hoards and the Pattern of Violence in the Late Republic’, PBSR 24 (1969), 79, shows a high frequency of unrecovered hoards in Italy between 75 and 71 B.C., but a low frequency 70–50 B.C. The evidence is suggestive only. I have suggested (n. 21) that the constant loss rate was very probably less than 10 per cent.

32 cf. Frederiksen, M. W., ‘Caesar, Cicero and the problem of debt’, JRS 56 (1966), 131Google Scholar ff.; cf. Crawford, M. H., ‘Money and Exchange in the Roman World’, JRS 60 (1970), 46;Google Scholar see also Cicero, Letters to Atticus 7. 18.

33 To some extent, this must have happened. And as a result, migrant labourers were attracted to places with high levels of expenditure, such as frontier garrisons (hence the urban development there) and to the city of Rome.

34 I started by re-analysing Dr. Richard Reece's data, published in ‘Roman Coinage in the western Empire’, Britannia 4 (1973), 227 ff.CrossRefGoogle Scholar I am most grateful to him for discussing his data with me, and particularly for reworking his data from northern Italy for the period A.D. 69–96 into two sub-periods, 69–81, 81–96. The patterns which emerged enticed me to see what I would find from other regions or coin collections.

35 In addition to Reece's data, I used six volumes of Gebhart, H. and Kraft, K. (edd.), Die Fundmünzen der römischen Zeit in Deutschland (1961–)Google Scholar: Saar, Pfalz, Süd-Baden, Süd-Württemberg, Schwaben, Oberbayern, i.e. a band of adjacent districts in southern Germany. For hoards in the Balkans and in Britain, Gaul and Germany, I used Bolin, S., State and Currency in the Roman Empire (1958)Google Scholar, appendices; from this collection, I arbitrarily excluded from consideration one enormous Bulgarian hoard of more than 60,000 silver coins, which overwhelmed the other finds, and which seemed different in character from the other hoards. Finally, I used Bellinger, A. R., The Excavations at Dura-Europus, Final Report VI, The Coins (1949)Google Scholar. I should note that the museum collections from Britain, northern Gaul (including some from northern Germany) and southern Gaul cover a large area. I checked before compression of the districts, separately analysed by Reece, to make sure that the patterns being compressed were roughly similar, so that the single line drawn from the collections in Figure 4 reasonably reflects the individual components.

36 In fact, following Reece, I collapsed short reigns with adjacent longer reigns (for example, Titus with Vespasian, Nerva with Trajan) to form convenient regnal periods.

37 For example, there were 3,812 singly-found silver coins listed in the six volumes of German coin finds, which I analysed. Incidentally, this was by far the smallest number of coins used for any line in Figure 4. The average number of coins per year in the period A.D. 96–180 was 5·3; 5·3 is the second-century index number (100) for south German singlyfound coins. For the reign of Hadrian, the average number of coins per year was 4·9 which is 92 per cent of the second-century index number, and so on.

38 Denarii and so-called antoniniani have been treated equally as silver coins. If we had taken account of the face value of antoniniani (at two denarii), which were minted in large quantities only after A.D. 238, then the lines at the right end of Figure 4 would have been higher than shown.

39 In addition to the problems of debasement, of reminting old coins and of loss-rates, which I have already mentioned, coin volume in the High Empire is complicated by the operation of several mints in the eastern Mediterranean. Thanks now to the painstaking and impressive work of Walker, D. R., The Metrology of the Roman Silver Coinage I–III, British Archaeological Reports, Supplementary Series 5, 22, 40 (1976–8Google Scholar), we can see how Roman provincial mints (for example, in Syria, Asia Minor and Crete) reflected a central Roman monetary policy. The evidence for this central control (which was directive, not reactive) is that the weight and fineness of provincial silver coins were reduced roughly to the same extent as, and sometimes before, silver coins minted in the city of Rome. This co-ordination of imperial monetary policy has important historical implications. But it was imperfect, so that measurement is difficult.

40 The problem is complicated. Figure 4 implies that the volume of debased silver coins rose considerably after A.D. 193 and that the increase in the volume of coins minted outstripped the rate of debasement. For example, I reckon from Walker's data (see previous note) that the median weight of silver in denarii minted in the city of Rome fell by 43 per cent at most, between A.D. 180/9 and 211/17 (2·29 g of silver in 180/9, 1·85 g in 196/211, and 1·31 g equivalent in the debased antoniniani minted at the end of Caracalla's reign (face value 2 denarii)). But the proportionate increase in the number of coins found is visibly greater than 43 per cent (see Figure 4). Such an increase in money supply might initially have stimulated commerce and production; but the increase in coins was too rapid not to have increased prices also. And in due course there was a downturn in trade; I am not claiming that increased money supply and price rises were alone responsible for the downturn in trade in the third century. Some of these issues are excellently discussed by Corbier, M., ‘Dévaluations et fiscalité (161–235)’, in Les dévaluations à Rome, Collection de l'école française de Rome 37 (1978), 273Google Scholar ff.; I disagree with Corbier in important detail, while admiring her work in general.

41 The most obvious index of urban decline is the widespread drop during the third century A.D. in the number of datable inscribed stones, commemorating the erection of new buildings, charitable foundations, statues, gifts, manumissions and deaths. See, for example, the statistical analysis of some evidence by Duncan-Jones, R. P., The Economy of the Roman Empire (1974), 352Google Scholar, and Laum, B., Stiftungen in der griechischen und römischen Welt (1914)Google Scholar. Changes in the fashion for inscribing and giving may account for some of this drop, but surely not for all of it. Yet how sensitive are such inscriptions as an index of prosperity, and of whose prosperity? For other illustrative evidence of urban decline see, for example, Duval, P.-M., Paris antique (1961), 277Google Scholar, and, for a fourth-century revival, Patlagean, E., Pauvreté économique et pauvreté sociale à Byzance (1977), 232CrossRefGoogle Scholar. I cite these isolated illustrations in the absence, as far as I know, of synoptic archaeological reviews.

42 The silver coin types listed, by obverse and reverse types, in the several volumes of Coins of the Roman Empire in the British Museum, ed. H. Mattingly et al. (1923–) are a tenuous index of the volume of coins ever minted, since we do not know how many identical dies of the same type were used. But no one can reasonably doubt the increase in the volume of silver coins minted in the Principate. This absolute growth is important, but once it is divided by the size of the population (coins per head) it becomes less impressive.

43 This is what I have called a wigwam argument, in which weak arguments prop each other up and circumscribe ‘truth’; see my Conquerors and Slaves (1978), 20.

44 See particularly Jones, A. H. M., The Roman Economy, ed. Brunt, P. A. (1974), especially 161 ff.Google Scholar; Marquardt, J., Römische Staatsverwaltung2 (1881)Google Scholar is still useful. Much more has been written about taxation in the Late Empire; see particularly Déleage, A., La capitation du Bas-Empire (1945)Google Scholar, and Cerati, A., Caractère annonaire et assiette de l'impôt foncier au Bas-Empire (1975)Google Scholar. But lengthening a bibliography should not disguise our lack of solid information and of real understanding about Roman taxation. Some Romans knew the size of their own state expenditure. Appian (Roman History, Preface 15) promised that, in his last book, he would outline the size of Roman military forces, the revenues collected from each province, the cost of the navy, etc. Unfortunately, this book does not survive. Augustus, the first emperor, left a will in which he detailed ‘the cost of the army, revenues, public expenditure, the amount of money in the treasuries …’ (Dio 56. 33). The conjunction army, revenues, expenditure is suggestive.

45 Revenues of the Roman State. The major problem in estimating Roman state revenues is the quality of the surviving evidence. The following six snippets have been trusted more than they deserve: (a) Julius Caesar imposed a tribute on Gaul of 40 million HS (Suetonius, Julius Caesar 25); (b) ‘(The Gauls) pay almost as much tribute into the treasury as the rest of the world’ (Velleius Paterculus 2. 39); (c) ‘Augustus made Egypt tributary, thereby contributing nearly as much revenue to the treasury as (Caesar) had brought to it from Gaul’ (ibid.); (d) Herod Agrippa derived from Palestine ‘as much revenue as possible, amounting to twelve million drachmae’ [= denarii] per year (Josephus, Jewish Antiquities 19. 352); (e)‘…. the tribute which (Egypt) yields to Rome in one month surpasses what you (in Palestine) pay in one year; besides money, (Egypt) sends wheat to feed Rome for months’ (Josephus, , Jewish War 2. 386)Google Scholar; (f) ‘When the cost had reached seven million (drachmae), the procurators of Asia wrote to the emperor that it was a scandal for the tribute of five hundred cities [i.e. Asia] to be spent on one city …’ (Philostratus, Lives of the Sophists 548).

There is not much more than this. Frank seeks to make (a) compatible with (b), and (c) compatible with (d) and (e), by claiming that they refer to different periods and that tax rates rose in Gaul and in Egypt between the beginning of the reign of Augustus and the middle of the first century A.D. But from 40 million HS to half the revenue of the empire in die case of Gaul? And from less than 40 million HS to more than 500 million HS (12 × 12 million denarii) in the case of Egypt? Yet Asia was reportedly paying less than 30 million HS in the second century A.D. (f) I That is absurd. See Frank, ESAR v, 7 and 51 and n. 49 below on the revenues of Egypt.

46 It is impracticable to calculate the total cost of the Roman army thoroughly in a footnote. It is a reflection on scholarly concern with detail, rather than with broad problems, that I can cite no standard estimates of how much the Roman army cost. My tentative conclusion is that the total cost of the Roman army at the beginning of the first century A.D. was 415 million HS, plus or minus 50 million HS. See Appendix I for details.

47 ‘Our present revenues are insufficient to provide for the army and everything else’, wrote Cassius Dio (52. 6) in a speech which he attributed to Agrippa in 29 B.C. Another writer (SHA, Probus 23) envisaged a dream world in which there would be no soldiers and therefore no tax on land. The jurist Ulpian explained that tax (tributum) was what was attributed to soldiers (D. 50. 16. 27: ‘sane appellatur … tributum … ex eo quod militibus tribuatur’). The sixth-century anonymous author of Practical Politics wrote that ‘expenditure on the army is the biggest item of state expenditure each year’, Griechische Kriegschriftsteller, edd. Köchly, H. and Rüstow, W. (Leipzig, 1855) 11.2, 47Google Scholar. None of these sources is earlier than the third century A.D.; indeed their citation may not be convincing. Yet it seems likely that their statements were broadly true, and that the cost of the army dominated the state budget during the High Empire also.

48 This calculation involves multiplying legionary pay by the average weight of, and by die proportion of silver in, coins minted in the city of Rome under each emperor. According to this calculation, legionary pay was for long periods under the Augustan level, and rose by more than 10 per cent above the Augustan level only briefly, in the reign of Caracalla. See Watson, G. R., The Roman Soldier (1969), 91Google Scholar on soldiers' pay, and D. R. Walker, op. cit. (n. 39), for the weight and silver content of coins. To be sure, in so far as coinage was fiduciary, i.e. in so far as its worth did not depend upon its silver content, such a calculation tells us little. I am willing to believe that in small-scale transactions, the silver currency was substantially fiduciary (since testing coins for exact silver content would have been difficult), but, in gross, I imagine that prices were influenced by the increases in money supply which followed debasement.

49 On the basis of two snippets from Josephus, quoted in n. 45 (d) and (e) above, Frank (ESAR v, 52) concluded that Egypt yielded 576 million HS plus 20 million modii of wheat (at say 3 HS per modius = 636 million HS total). The population of Roman Egypt is conventionally regarded as above seven million (Finley, The Ancient Economy (1973), 97); in K. J. Beloch's view (and I agree completely) that is far too high (Die Bevölkerung der griechisch-römischen Welt (1886), 258 and 507: 5 million); even Beloch's estimate is generous. Seven million people would have been paying 200 kg wheat equivalent per person in tax to yield Frank's estimated total. That is again much too high: the claimed total is five times higher in wheat terms than the total tax levied in Egypt by the Ottomans in the seventeenth century (see S. B. Shaw, The Financial and Administrative Organization and Development of Ottoman Egypt (1962), 79, 84 and 183. Cf. the implied tax rate in the fourth century A.D. (P. Oxy. 3307) which was much lower, and n. 56 below.

50 Russell, J. C., ‘Late Ancient and Medieval Population’, Transactions of the American Philosophical Society 48. 3 (1958), 7CrossRefGoogle Scholar, for literature and for a full discussion of the evidence, and see n. 52 below.

51 Clark, C. and Haswell, M., The Economics of Subsistence Agriculture4 (1970), 57 ff.CrossRefGoogle Scholar and 175. I once did some fancy calculations allowing for body weight (adult males 60 kg at age 25 years), age structure (e0 = 20), climate (at Rome), subsistence at 2,000 calories average per person/day. The result coincided with Clark and Haswell's. I added a bit (15 kg wheat equivalent per person/year) for clothing, and a similar notional amount for heat and housing. The end result (250 kg wheat equivalent per person/year) is obviously rough and speculative (after all consumption depends on energy expended and vice versa); I thought it best to express the result in a round number (250 kg) to underline its vagueness. But the probable margin of error is not great.

52 Beloch, op. cit. (n. 49), 507 and ‘Die Bevölkerung im Altertum’, Zeitschrift für Sozialwissenschaft 2 (1899), 505Google Scholar ff. and 600 ff.

53 Columella wrote: ‘We can hardly remember a time when cereals in the greater part of Italy yielded four to one.’ On varying yields in one Italian district, see, for example, Rotelli, C., ‘Rendimenti e produzione agricola nell'Imolese’, Riv. Stor. Ital. 80 (1968), 121–3,Google Scholar and Aymard, M., ‘L'agriculture dans l'ltalie moderne’, Annales 28 (1973), 475–97CrossRefGoogle Scholar.

54 See M. I. Rostovtzeff, RE s.v. frumentum, 149; cf. Duncan-Jones, R. P., ‘The Price of Wheat in Roman Egypt under the Principate’, Chiron 6 (1976), 251–3,Google Scholar who lists eighteen wheat prices from lower Egypt in the first century A.D.; the median and modal price was 8 drachmae per artaba of 32 kg, which Duncan-Jones approximates to 2½ HS per modius. By the end of the second century A.D., lower Egyptian wheat prices had more than doubled to 18–20 drachmae per artaba, but only four prices are known and exactly dated A.D. 191–220. In Asia Minor at the end of the first century A.D., in a small town (Antioch in Pisidia), the normal price of wheat was 2½ HS per modius (AE 1925, 126). In the city of Rome, market prices were obviously higher, perhaps 8–10 HS per modius (cf. R. P. Duncan-Jones, The Economy of the Roman Empire (1974), 345–7). Prices fluctuated both within and between years; fluctuations do not preclude an average, but they should induce caution in its use.

55 No attempt was made to impose a uniform tax-system or a single tax-rate on crops and land throughout the empire, though some taxes (inheritance tax on citizens, customs dues) were raised across the empire. Some lands in Roman Germany were called agri decumates, tithe lands, and tithes had been raised during the Late Republic in Sicily and Asia Minor, by tax-farmers. But Hyginus, who wrote in the early second century A.D., mentioned tax-rates of one-fifth and one-seventh (ed. K. Lachmann (Berlin, 1848), 205); these rates may have been due to local variations, or the result of a rise in tax-rates (perhaps under Vespasian, see n. 68). In Syria, the tax-rate was 1 per cent of the assessed value of the land (Appian, The Syrian Wars 50). These were the main taxes, to which we should add indirect taxes. But for the moment I am taking no account of illegal exactions and squeezes. I am concerned only with what the central government and its agents took officially, in whatever form and wherever spent. I have not touched the problem of how and when a tithe of a main crop, such as wheat, was transformed into money. I can only stress the great difference between a declared tax-rate (say 10 per cent) on a main crop, and my first estimate of government revenues as 10 per cent of all produce. The survey by Cuinet, V., La Turquie d'Asie (1891) I–IVGoogle Scholar, gives a detailed analysis of taxes raised in Asia Minor and Syria by the Ottomans at the end of the nineteenth century, and suggests the different contributions from land taxes, cattle taxes, customs which could be raised in a still undeveloped economy.

56 I shall first state the results, then the elements in the calculation, then the sources. Results: State budget expenditure per head of population (in kg wheat equivalent)

Sources: United Kingdom: Mitchell, B. R., Abstract of British Historical Statistics (1962), 5, 386–91, 486–8Google Scholar; Chandaman, C. D., The English Public Revenue 1660–1688 (1975), 208Google Scholar; Cipolla, C. M., Before the Industrial Revolution (1976), 4Google Scholar. France: Braudel, F., The Mediterranean (1972) I, 395–421Google Scholar; R. Baehrel, op. cit. (n. 4), 535; H. See, Histoire économique de la France (1948) I, 157 ff.; 11, 111–22; C. M. Cipolla, op. cit., 4; E. A. Wrigley, Population and History (1969), 153. Needless to say, the results are crude and should be treated with the utmost caution.

57 Brunt, P. A., ‘Charges of Provincial Maladministration under the early Principate’, Historia 10 (1961), 189 ff.Google Scholar

58 McKnight, B. E., Village and Bureaucracy in Southern Sung China (1971), 7Google Scholar; cf. Twitchett, D., Financial Administration under the T'ang2 (1963), and 217Google Scholar.

59 The system is clear from IG v (1), 1432–3, convincingly dated by Giovannini, A. (Rome et la circulation monétaire en Grèce (1978), 115 ff.)Google Scholar to A.D. 35–44. My interpretation of this important inscription is that the Romans levied a tax of 100,000 denarii on the town of Messene in southern Greece. The town then divided the tax due by the total declared capital value of property including agricultural holdings (which the inscription lists by district totals) and thereby arrived at a tax-rate, so much per 100 drachmae or denarii (in fact 8 obols = 1·3 per cent). It is noteworthy that outsiders, xenoi explicitly including Romans, had the highest rate of non-payment, at the time the inscription was carved. Cf. A. Wilhelm, JOAI 17 (1914), 1–120, for a detailed and interesting commentary with which I reluctantly disagree in part.

60 I imagine such techniques as collusively low valuation on the élite's own property, early collection of other people's taxes, and late payment of taxes by the rich; loans by the rich to the poor against the surety of their land. My main appeal is to the logic of the situation, and to comparable data from other societies (W. Hinton, Fanshen (1966), 39; Spence, J. D., The Death of Woman Wang (1978), 43 ff.Google Scholar; Huang, R., Taxation and Governmental Finance in Sixteenth Century China (1974)Google Scholar. But Roman evidence also exists; see, for example, Jones, A. H. M., The Later Roman Empire (1964), 467–9Google Scholar.

61 A famous plea survives from the tenants of an imperial estate in North Africa; they had already appealed to the emperor's local agent (procurator), but he was in cahoots with the administrator or lessor (conductor) of the estate: ‘… a collusion which he has practised uninterruptedly not only with Allius Maximus, our oppressor, but also with almost all the lessors, against the law, to the detriment of your treasury. The result is that he has refrained from investigating, for many years, our petitions, supplications and our appeals to your divine rescript; more than that he has yielded to the wiles of the said Allius Maximus, lessor, … to such an extent that he has sent soldiers into (our estate) and given orders that some of us be seized and tortured, and others … be beaten with rods and cudgels although they are Roman citizens’ (ESAR IV, 98 = CIL VIII.10570, cf. 25902 and 25943).

62 The logic of the situation and comparative evidence both suggest what we should expect. The explicit recognition of this tactic in research violates the implicit rule or convention among ancient historians that the surviving testimony provides both the building bricks for our history and its authentication. But by what logic do we decide whether the surviving testimony is true, or representative, and how do we decide between conflicting sources? These are not just problems of historical philosophy; they are recurrent problems of historical interpretation.

63 Not that the surplus was fixed in size. Indeed, the imposition of money taxes and rents probably made peasants increase the size of the surplus produced. But the potential for growth was narrowly finite. Private profit therefore competed with public exactions. I should stress that the concept surplus is ‘objective’: what was produced over and above minimum subsistence. Peasants may have wanted to consume it themselves; they probably did not regard it as surplus to their needs.

64 Tax-farmers' charges presumably reflected their administrative costs, plus their risks, plus their interest charges on the capital which had been advanced to the Roman government, plus overcharging (loss). E. Badian, Publicans and Sinners (1972), in a sympathetic account, rightly stresses how difficult it would have been for the Roman state to administer its large new empire without private entrepreneurial help.

65 Early in the reign of Vespasian, the senate voted to accept a loan of sixty million HS from individuals, but it was never taken up (Tacitus, , Histories 4. 47)Google Scholar.

66 Senators in the early second century A.D. were formally required to hold one-third of their fortunes in Italian land (Pliny, , Letters 6. 19)Google Scholar; the proportion was later reduced to one-quarter (SHA, Marcus Aurelius 11).

67 According to Pareto's law, the proportion of total wealth held by the wealthy minority in preindustrial states is constant (see Pareto, V., Cours d'économie politique (1897), 964)Google Scholar. Subsequent research has cast some doubt on the strict universality of the law. But it remains suggestive. We should expect the total wealth of Roman senators and knights to grow commensurately with the growth in the size and wealth of the empire.

68 So Jones, A. H. M., The Roman Economy (n. 44), 177Google Scholar; Suetonius, Vespasian 16: ‘he increased tribute from the provinces’; and see n. 55 above. Silence is of course not proof.

69 Emperors and their advisers do not seem to have realized the consequences of their repeated decisions to debase coins. But then in post-feudal Europe, when the consequences of debasement were roughly known, debasements still occurred, because of their short-term advantages. Cf. C. E. Challis, Tudor Coinage (1979); on Roman debasements see Callu, J.-P., La politique monétaire. des empereurs romains de 238 à 311 (1969)Google Scholar and A. H. M. Jones, op. cit., 187ff.

70 Mickwitz, G., Geld und Wirtschaft im römischen Reich (1932), 120Google Scholar shows that the proportion of Egyptian land rents (N = 301) expressed only in natural produce rose considerably in the fourth century A.D. compared with previous centuries. And in northern Italy ritual fines for violators of graves, threatened on tombstone inscriptions, were in the early fourth century A.D. expressed in weights of silver and gold, instead of in coin as previously (CIL v, 8721 ff.); on which see Pekáry, T., ‘Studien zur römischen Währungs- und Finanzgeschichte’, Historia 8 (1959), 462Google Scholar.

71 P. Oxy. 1411; cf. Rostovtzeff, M. I., Social and Economic History of the Roman Empire2 (1957), 470 ff.Google Scholar

72 For theoretical and comparative works, see Ardant, op. cit. (n. 28); R. M. Bird, Taxing Agricultural Land in Developing Countries (1974); see also Cheung, S. N. S., ‘Private Property Rights and Share Cropping’, J. Pol. Econ. 76 (1968), 1107–22Google Scholar, and Issawi, C., ‘Farm Output under Fixed Rents and Share Tenancy’, Land Economics 33 (1957), 74–7CrossRefGoogle Scholar, for similar problems in relation to rent. The most sophisticated ancient discussion of taxation is in the speech attributed to 29 B.C. but written in the early third century by Cassius Dio (52. 28 ff.). One should also note the early Arabic treatises on taxation, dating from the eighth century A.D. onwards, which probably in part derived from lost Byzantine texts or from Byzantine practice. See Ya'qub ibn Ibrahim alias Abu Yousouf Ya'qoub, Le livre de l'impôt foncier, trans. E. Fagnan (1921), esp. 74–5, and Shemesh, A. Ben (ed.), Taxation in Islam (1969) III, 100–1Google Scholar.