This article is available at the URI as part of the NYU Library's Ancient World Digital Library in partnership with the Institute for the Study of the Ancient World (ISAW). More information about ISAW Papers is available on the ISAW website.

©2019 Roger S. Bagnall and Gilles Bransbourg; text and figures distributed under the terms of the Creative Commons Attribution 4.0 International (CC-BY) license. Coin images in the plates are ©2019 American Numismatic Society and distributed under the Creative Commons Attribution No Commercial 4.0 International 4.0 (CC-BY-NC) license.
Creative Commons License

This article can be downloaded as a single file

ISAW Papers 14 (2019)

The Constantian Monetary Revolution

Roger S. Bagnall and Gilles Bransbourg

Abstract: The fourth century CE represents a peculiar moment of monetary history. Most prices rose about fifty thousandfold, materializing the strongest inflationary period ever experienced during Antiquity. Traditionally, this price inflation has been linked to coinage debasement. However, the reality is more complex: imperial authorities also manipulated coinage tariffs in current units of account. This is particularly noticeable under the reign of Constantius II, when most prices increased about twenty-fold in a matter of few years in the early 350s, with no coinage change of comparable magnitude. Very interestingly, gold and silver rose to preeminence at the same moment, at the expense of base metal. We believe both phenomena were linked. A thorough analysis of papyrological and numismatic evidence will demonstrate that the increased supply of silver coinage was allowed by the removal of silver from the existing billon coinage supply, while growth in gold coinage depended on new metallic sources. The sudden price increase, sometimes explained by some form of competition between precious and base metals, would in fact result almost mechanically from the retariffication and subsequent demonetization of the existing billon coinage, replaced during that process by bronze coins of comparable monetary value but of much lesser commodity value. This led ultimately to the bimetallic gold:bronze bullion-based price system that defines the Byzantine period. This paper originated in a conference presentation at "Money Rules!", held in Orléans October 29-31, 2015, and organized by Thomas Faucher. A slightly different version will appear in the proceedings of that conference: R. Bagnall and G. Bransbourg, The Constantian Monetary Revolution. In Th. Faucher (ed.), Money Rules! The monetary economy of Egypt, from Persians until the beginning of Islam (Cairo, IFAO, forthcoming).

Library of Congress Subjects: Numismatics, Roman; Economic history--To 500.


1. The Evidence of the Documents

To both numismatist and papyrologist, the evidence indicates major changes in the use of money in Egypt just after the middle of the fourth century CE, during the reign of Constantius II (337-361). In what follows we set out this evidence —a relatively easy task despite the gaps in the record— and try to assess its significance—a much harder affair.

The central role of gold in late antique Egypt has been widely recognized and debated.1 The vehicle of this role was the four-gram solidus introduced under Constantine and ubiquitous in the papyrological documentation.2 But this coin was far from having an immediate impact; as was observed thirty years ago, the attestations in the papyri almost all date after 350.3 There are even now just five instances known to us of mentions of solidi in the papyri before 350; all five are references to prices at which the coins could be purchased, none a transaction in which the solidus is used to pay for anything.4 In fact, there are even entries for purchases of a half solidus and two-thirds of a solidus, although these correspond to no numismatic reality at this period. These attestations are few: they amount to about one per every 75 published papyrological documents dating from the period 300-350.5

The contrast with the following half-century is marked. Nomismatia are mentioned in 38 texts, or once for each 2.9 papyri. And a property transaction of 27 March 352 (P.Stras. 1.9) is the first we know of to specify gold coinage.6 The rupture with past practice is sharp. Although the use of coinage largely composed of copper, denominated in either the Greek system of talents and drachmas or the Roman units of denarii continues throughout Roman rule, gold becomes dominant in larger transactions from this point on.

With prices denominated in the traditional units of account (drachmas, talents, denarii), which we may assume were normally paid in the standard billon nummi, the years just after 351 also mark a sharp divide after a half-century of evolution.   That half-century saw what has often been described as inflation, but was in fact, as showed in Currency and Inflation (Bagnall 1985) instead a series of steps in which the metallic value of the everyday coinage was reduced, partly by a reduction in the amount of silver in the coins and partly by a decline in the overall weight of the coins. Each such step led to a noticeable rise in prices.7 Where the stated denomination of the nummus remained the same after one of these reductions, it had the effect of raising the market value of the metals in the coin in nominal terms; when the nummus was revalued to a higher amount, that rise in market value was exaggerated. Other prices followed suit. Even reforms that improved the quality of the coin could have such effects, if they involved retariffing the coin at a higher value.

Chart 1: Price of gold, 275-330. After Carlà 2009, Grafico 1.
Chart 2: Price of gold, 330-360. After Carlà 2009, Grafico 2.

The results can be shown in graphic form. We reproduce here two such charts from Carlà 2009: 29-30. These show the rise in the nominal price of gold bullion between 275 and 360. These charts in one sense misrepresent the phenomenon by the use of continuous lines, as if a constant rate of “inflation” was at stake, when in fact we are dealing with a step function.8 But they do give a good sense of the incrementalism visible in most of the coinage reductions, as well as a sense of the exceptional place occupied by the “reform” of 325 and the Constantian revolution under discussion here.

The argument for this view of the price changes visible in the papyri and their relationship to the coinage has been widely accepted by both numismatists and papyrologists, although not without some controversy, at least initially; it was at the center of much of the discussion of a conference held in Rome in 1988.9 It should, however, be stressed that some caution is in order in dealing with the dates of texts published after 1985, because the discussion of prices in Currency and Inflation has had a profound impact on the later papyrological literature, where dates have often been assigned on the basis of the chronology outlined there. This is quite reasonable, but one must remain alert to the distinction between demonstrable and assigned dates and the risk of circular argument.

The caesura of 351 in the prices reflected in the papyri is somewhat less easy to measure and describe than the sudden rise of occurrences of prices given in gold mentioned earlier, because it is not a yes-no binary but a matter of scale, and much of the relevant evidence has no precise date. Again, the outlines of this shift were described in 1985 in Currency and Inflation. A certain amount of new evidence had come to light by 1997, when a supplementary list of prices was published in the appendix to the Kellis Agricultural Account Book (Bagnall 1997), but not much new has been published since then, nor are we aware of substantial bodies of documents awaiting publication.

The overall nature of the change can be seen by comparing the price of wheat found in P.Oxy. 1.85, securely dated to 338, of 24 talents per artaba, with that in SB 14.12154, firmly dated to 357/8, of 846 talents per artaba. A 35-fold rise in prices within a twenty year period is formidable. The lack of firmly datable commodity prices in the intervening years complicates our task, but the rise is so dramatic that even non-standardized objects of transactions and approximately datable documents are useful in narrowing down what happened. These suggest that prices rose only moderately during the 340s, through the period that includes most of the texts of the Abinnaeus archive (Bell et al. 1962).

Looking at housing costs turns out to be very helpful, despite the fact that houses obviously vary greatly in size, location, and quality. We have several securely dated examples from the 340s:

Slightly later, we have PSI 6.707, dated to February-March 351, where the rent is 150 talents/year for a half house. If we imagine that this half house included several rooms, the figure does not seem out of line with the rents for the carpetmaker or the basement room. It is certainly more than the Harris papyrus with its entire house, but it would not be hard to find in any time a similar range in house rents. Clearly on the other side of a divide is P.Abinn. 22, where a house (in Alexandria, to be sure) rents for nearly 1,550 talents per year. Unfortunately, this letter has no precise date. The latest datable texts from the Abinnaeus archive are from 351, so it is unlikely that the letter dates from much after that year; but it is entirely possible that some of the undated texts from the archive belong to the years immediately after 351. Somewhat harder to parse is P.Dubl. 31, a lease of a linen-weaving workshop in Panopolis dated to 355, where the rent is 200 talents per year. As this is a single room, one imagines, it is perhaps to be compared with the 20 talents for the carpenter’s space in 344. A ten-fold increase would be indicated by this comparison. Later rent rates, dated 360 and on, are in the high hundreds to thousands.10

Pinning down the timing of the change in price levels is not easy, because the 350s are singularly lacking in datable prices for standard commodities. The most useful price is that for the rent on a palm grove recorded in P.Oxy. 14.1632, which was already cited in Currency and Inflation as the earliest figure clearly belonging to the new regime. In this text, dated 25 July 353, the crop for a single year for two choria (garden plots) is leased for a rent of 8,000 talents. Before the great divide, this would have been something like the value of 200 artabas of wheat, the rent on a substantial estate. Although the size of choria could vary, they are mostly only an aroura or two in size, and this seems like an impossible rent at the pre-351 price levels.

It is frustrating that the last thirty years have brought no new price information for the early 350s that would help us make the nature of the transition more precise. It was already clear then that the commodity prices known from around 360 implied a gold price of somewhere between a half-million and a million talents per pound. The silver price in P.Oxy. 51.3624, securely dated to 359, and the wheat price in P.Oxy. 51.3625, of the same year, would point to the lower half of this range as prevalent at the end of the 350s. But we know that the initial jump is likely to have been more like 10-fold than 20-fold from the gold price around 30,000 talents/pound prevalent in the 340s. For example, the undated P.Princ. 3.188v has wheat at 334 talents/artaba, compared to the 24 talents per artaba in P.Oxy. 1.85, securely dated to 338; P.Oxy. 34.2729 shows gold at 350,400 talents a pound. And the reediting of SB 16.12828 has now shown that it uses consistently a talents per pound of gold equivalence of 276,480. None of these can be dated independently of their prices, but all must be later than February, 351—but perhaps not much later.

To sum up: A sharp caesura in prices denominated in talents and drachmas, or alternatively in denarii (the denarius was equal to 4 drachmas) seems to occur between February 351 and July 353. It cannot be pinned down more precisely with present evidence. It is argued already in Currency and Inflation that the term myriad (i.e., 10,000), used without explicit specification of drachmas,11 talents, or (most usually) denarii, but in fact referring to denarii, comes into use in connection with this caesura. There is in fact not a single example of this usage of myriad by itself to refer to denarii that is securely datable before 351. There are two instances in the Abinnaeus archive that lack precise dates. One is the letter referring to house rent already mentioned, P.Abinn. 22, which seems to belong after the price rise. The other is P.Abinn. 81, an undated account largely listing items of clothing, with prices ranging from 58 to 72 myriads, and hides at 30 myriads. One may compare P.Oxy. 12.1431, with its figure of 225 myriads (of denarii, specified, and equated to 1,500 talents) for a rug for the visit of the dux. It is pretty difficult to know how to evaluate this figure, which is to be dated early in 351; possibly 3 January, but only two fragmentary letters are preserved of the month name. At all events, the Abinnaeus account would be unique in its use of myriads tout court if it dates before 351. This is just one more pointer to the abrupt nature of the change at stake.

2. The coinage

2.1. Coinage from 294 to 351

Fourth-century Roman coinage remains a challenging numismatic field. With the dislocation of the classic Roman imperial coinage during the third-century monetary crisis, the denarius moved toward the status of a ghost currency, a unit of account diverging more and more from the physical coinage as we progress through the century. Since marks of value were normally absent from the coins, attributing currency values to each physical coin type sometimes leads at best to educated guesses.

Diocletian reformed the coinage system starting in 294 (the date varies by mint), aiming at reintroducing a stable monetary system inspired by the Augustan trimetallic coinage structure of the earlier Imperial period. A gold coinage whose weight standard settled at 1/60 of a Roman pound (ca. 5.40 g) aureus and a 1/96 lb. coin, marked XCVI to guarantee its weight, with a high silver content (ca. 92%) argenteus were minted (Plates 1 and 2). They emulated the High Empire aurei and denarii struck respectively at 1/44 and 1/96 lb. under Nero after the reform of 64. Its implementation had to wait until 296 in Egypt and was followed by a brief period of uncertainty during the usurpation of L. Domitius Domitianus in 297-298.12

The two most original parts of the new system lay with the use of the denarius as a common unit of account for the entire coinage system and with the introduction of a large laureate billon nummus weighing ca. 10 g (1/32 lb.) and containing around 4% of silver (Plate 3). The coin was tariffed at 12.5 denarii before the currency reform of 301 CE, while the argenteus was probably worth 50 denarii, and the 1/60 lb. aureus (which is never tariffed explicitly by the surviving documentation and might have fluctuated if one follows F. Carlà) traded most likely at 1,000 denarii at that time, since a pound of gold was valued 40 talents (= 60,000 denarii) in an official transaction in 300 CE.13  As a result, their respective values would have stood at 1: 20: 80, quite close to the 1: 25: 100 ratio that had linked the aureus, denarius and sestertius of the earlier period. Small radiate and laureate billon and bronze coins that had circulated since Aurelian completed the system as far as low denominations were concerned. (Plates 4 and 5). Assuming from Diocletian’s 301 CE Price Edict a 1: 12: 1,440 ratio between the values of gold, silver and bronze,14 the silver coin would have been slightly undervalued vs. the gold coin, by about 10%, while the billon coin would have been overvalued by almost 50% before the currency reform of 301 CE-- quite in line with the practice of the earlier Empire. Copper radiates and laureates (1.3 g for the latter) were minted as coins of low value and embodied the denarius with values in a 1-5 denarii range.

The reason why the main piece of the system, the billon nummus, incorporated silver remains to be explained. Under Augustus, the role of a coin intermediate between silver and copper was supplied by the use of orichalcum, an alloy of copper and zinc more valuable than bronze. Based on the 301 CE Price Edict, where orichalcum as a raw metal is valued at a ratio of 1: 720 vs gold, a sestertius-type 1/9 – 1/12 lb. piece (ca. 30 g) could have played the same role in the currency system with the same degree of relative fiduciarity as the 10 g billon nummus. Diocletian’s reformers instead obviously opted for a lighter coin with a small amount of silver. The authorities might have used silver as an easier manner of gaining some minting profit, as the XX-I (or KA) mark, suggesting a claim of 20 parts of copper for 1 part of silver, stands above the proportions of silver that have been generally measured, which are closer to 3%. Maybe zinc was in shorter supply. Another more fundamental reason may be found in the earlier Aurelianic reform of 274 CE, where the reintroduction of silver into a hopelessly debased billon coinage had marked the reconstruction of a non-inflationary system by lending credibility to the currency. In 16th – 17th century Europe, the replacement of billon by copper for small change coinage was effectively received with great reluctance by the public even when intrinsic values were comparable.15

The Currency Edict of 301 CE revalued the nummus to 25 denarii, and probably the argenteus was set at 100 denarii,16 while the price of the pound of gold in the Price Edict implies a 1,200 denarii price for the 1/60 lb. gold aureus - for which no tariff is provided. It is not clear whether or not the copper radiates and laureates were revalued, and their values remain unclear until now. The smallest piece, the laureate, is often assumed to have been worth 2 denarii whether before and after the reform, if it remained untouched, or after the reform if it was revalued as well.17

At that point, the silver piece was significantly overvalued, while the billon piece, the most produced coinage of the reform,18 enjoyed a metallic coverage standing at significantly less than half of its monetary value.19 Most prices must have jumped in the wake of the reform, hence the almost simultaneous Price Edict that follows –guaranteeing the purchasing power of the billon coinage, protecting the main holders of nummi, soldiers and imperial officials to some extant, and foremost the State through the system of compulsory requisition of gold and silver purchased at fixed rates against the billon coinage.20

The Tetrarchic years then witnessed a progressive debasement of the billon coinage through step-by-step weight and fineness reductions, while the coins’ monetary value remained unchanged in denarii. By 312-315, the nummus had lost more than 50% of its metallic content – weight and silver fineness combined. By 318-321, the standard piece weighed around 3.30 g with about 3% of silver, implying a relative loss of about 75% of its original metallic value.21 (Plate 6). In taking into account the two-fold increase in the currency value of the nummus in 301, one would expect approximately an eight-fold increase in most prices over the same period, provided users had tended to reckon the worth of the billon coinage at its commodity value during this debasement phase.

However, assessing the impact on overall prices is not a straightforward exercise, as most gold and wheat prices from these years involve official requisitions or forced conversions, while other prices often lack a comparative basis or may imply produce of varying quality – e.g., wine – not to mention possible seasonal variations. That said, prices seem to follow the metallic depreciation in rough terms. Two market transactions of gold in the 310-318 chronological range display a seven-fold increase in the price compared with 300 (P.Ryl.4.643 and P.Oxy.48.3121), a compatible order of magnitude.22 By enforcing a certain number of regulated and undervalued prices for official transactions, notably precious metal requisitions and grain adaeration (forced conversion of tax dues from cash into grain at official tariff), the state benefitted from the billon coinage’s loss of metallic quality. Two official prices from 324 materialize this policy, as gold is purchased at 168 and 209 talents per lb., while a private transaction from most likely 318 reveals a price of 288 talents, even though the coinage of 324 was much debased compared to that of 318.23

Indeed, very likely in 324, Licinius dramatically reduced the coinage’s commodity value and issued a ca. 3-3.30 g, 12.5 denarii piece marked XIIΓ that contains almost no silver – about 0.12%. Rebased in bronze, this is not even equivalent to 4 g all-in with a 120: 1 value ratio between bronze and silver. As the original Tetrarchic nummus was worth almost 60 g of equivalent-bronze, this is a 93.5% loss of metallic value, which would imply prices at least 15-fold higher than in 300. The two official gold prices from mid-324 under Licinius would imply only a 4 to 5-fold increase and we lack private prices for that year. Constantine demonetized this coinage in the fall of 324.24

The next step involved a currency reform, with a unique nummus type coined under Constantine between 324 and 330, struck on a 1/96 lb. standard (= 3.30 g) and incorporating about 2% of silver. Gold prices from the 325-330 period jump almost ten-fold compared to the market transaction prices from 318-323. As the billon coin from that period is quite comparable to its predecessor from the years 318-324, being debased only by about one-third, the only possible scenario involves a revaluation of the nummus. What probably occurred is the following: re-imposing pre-Licinius prices would have involved actual deflation, a result hard to achieve in the context of a general lack of confidence in the coinage. Post-325 billon coins must then have been valued at 100 denarii, hence the term centenionalis communis encountered in an imperial edict of 354 or 356 (C. Th. IX, 23, 1) – which likely does not refer to the 325-330 coins but at least shows that 100 denarii had become a standard value for the main billon coinage at some point.

There is not much to report for the subsequent period 330-342. The standard billon coin degenerates progressively into a 1.64 g 1.4% silver coin, and current prices move more or less in line with the base coinage commodity value. As we have few market prices for the later 340s, we finally reach the 348 monetary reform with the 348-352 fel. temp. reparatio series and the introduction of three denominations alongside a small number of standardized types: a large AE2 weighing over 5 g normally, a target weight of 1/60 of a Roman lb with a relatively high 2-2.5% silver content (Falling Horseman or Galley and Phoenix/Victory types), a reduced AE2 weighing below 5 g, a likely target weight of 1/72 of a lb, and incorporating less than 1% of silver (Hut/Barbarians or Captives types), and an AE3 of the Phoenix or Galley type of ca. 2.68 g – possibly 1/120 of a lb - with ca. 0.2% of silver.25  The combination of size, weight and symbols must have made clear to the users what denomination they were handling, the only potential exception being the dual use of the “Galley” type by Constans for both the large AE2 and the AE3. But the modules are very different, which must have ensured lack of confusion, while the frequent addition of letters contributing to the readability of the system.. (Plates 7-12) The AE2 coinage in the West sometimes bears lettered marks: A for the large coin (and sometimes Γ in the East), followed by B and Γ under Magnentius in the West (350-354), N for the smaller AE2, with Δ in the East, most likely marks of denominations. The AE3 stands as the most likely contender for the title of centenionalis by that time, which does not mean that it was tariffed at 100 denarii necessarily.26 The large AE2 is probably the pecunia maiorina referred to in 349, maintained until 351-352 with some early coins of Gallus (351-354), losing weight afterward, and demonetized before 354-356.27. Lack of securely dated commodity prices in the 340s, as described earlier, combined with uncertainties with respect to the value of copper vs. silver and gold, makes it hard to tariff with any confidence the AE3 piece, valued between 100 and 250 denarii by modern scholars.28 Some degree of confusion was brought by Magnentius’s usurpation in early 350, his coinage lost silver fineness and a heavier type was produced. Finally, the Falling Horseman AE2 remained almost the sole large billon type produced in the East at some point after the proclamation of Gallus in 351 and then for a brief period in the West after Magnentius’s fall in August 353, still marked A, albeit on a reduced standard close to that of the small AE2 (1/72 lb.), materialized by the LXXII weight mark displayed on the reverse by some of the series minted in Siscia possibly as early as 351, and then Aquileia after its conquest by Constantius II. (Plate 13). B and then Δ replaced the A behind the imperial bust in Rome, the same phenomenon (without the B) being visible in Siscia and Sirmium, highlighting the effective replacement of the large AE2 by its smaller version.29

The 348 reform can be summarized and simplified until 351 as follows for the most used types – excluding Magnentius’s types:

Table 1: Main reverse types of the fel. temp. reparatio

2.2 The 353-354 monetary change

Part 1 of this paper summarized the evidence that showed that a clear discontinuity separates Egyptian prices on both sides of a 351-353 period. The multiplier between the highest pre-351 and the lowest post-353 indexed prices stands at approximately fifteen, while a factor of about thirty-five separates the last firmly dated pre-351 evidence in 338 and the first securely dated post-353 evidence in 357/8.30 Then post-353 papyrological evidence points toward the existence of AE standard coins worth 10,000 denarii (the myriad) while later imperial edicts imply the existence of a coin named as a follis and valued at 12,500 denarii,31 in either case a huge jump from the previous 100-250 denarii rates.

The numismatic side does not display physically anything close to that discontinuity: the AE2 is effectively discontinued, replaced by a 2.5 g Falling Horseman AE3 with an improved ca. 0.6% of silver and the sporadic use of the letter M – potentially a mark of value. Since reduced AE2 are minted in Gaul for Constantius II and Gallus following the fall of Maxentius in 353, and disappear from the next series in 354, the AE3 would have taken over from the larger modules in early 354 in the West.32 However, nothing prevents that replacement to have occurred slightly earlier in the East. This AE3 type does not break with previous coinages from a metrological point of view – particularly the previous AE3 Phoenix/Galley type - to the point of justifying such a jump in current price levels. (Plate 14).

Then comes evidence of a rare demonetization, as C. Th. IX, 23, 1 (354 or 356) implies that a range of coins is by then forbidden and that everyone knows about it – the maiorina, the centenionales communes, and ceteras vetitas.33 Since the actual demonetization must have preceded such an edict, we have a clear chronological compatibility between the 351-353 price increase, the replacement of the AE2/AE3 348-352 fel. temp. reparatio series by the new Falling Horseman AE3,34 and the implication in 354 or 356 that some coinage demonetization had taken place not long ago. Kent suggested that the demonetization was more brutal in Gaul and Britain because of the local surge in forgeries at that time, as the new AE3 ‘effectively forced out of circulation all issues struck between 348 and 354, except the AE3 “Phoenix” type, which was metallically comparable to the later “Falling Horseman” issues.’35

It is natural to look to hoards to supply evidence of such a coinage replacement, particularly those from Egypt, since this is where the price evidence comes from. Three hoards indeed fit with a 353 discontinuity.36 Two other hoards – Kent reports a total of six hoards in Egypt for the 337-364 period – potentially cross the 353-354 divider: #206 ‘Mattingly’ and #211 ‘Milne’.

The ‘Mattingly’ hoard incorporates 1,484 coins and represents only part of the original hoard. The coins spread in the main from 349 until 354, as there are very few coins bearing the name of Constans (who died in 350) and no coins bearing that of Julian (Caesar in 355).37 Out of the 835 Alexandrian coins, where one would expect the most recent coins to show up, only 8 bear the M//ALEA mint mark, attributed by Kent to the first group of the 355-361 period. In fact, such a hoard would be typical of selected coins being voluntarily removed from circulation, possibly to avoid exchanging them at an unfavorable official rate. Of the 835 Alexandrian coins, 474 are AE2 (57%), as are as 81% of the 133 coins from Antioch, 78% of the 144 coins from Constantinople, and 80% of the 153 coins from Cyzicus, these being the most heavily represented mints, which seems to imply a preference for the largest denomination.

Unfortunately, the ‘Milne’ hoard is poorly documented.38 Kent put 362 as an end-date, based on some Julian Augustus series. Of the 650 coins, 202 are illegible. The AE2/AE3 split is missing from the hoard report. Out of the 448 identified coins, only 16 predate the fel. temp. reparatio introduction in 348. The number of coins bearing the name of Gallus (murdered in the winter of 354) is very low – 9. It is therefore almost certain that the hoard belongs essentially to the post-354 period.

As the sixth and last Egyptian hoard published by Kent ends before 348, it appears that none of the six hoards he documented from Egypt for the 337-364 period crosses the 351-354 divider in a significant fashion, lending credence to the effectiveness of the coinage demonetization that occurred during that period.39

Evidence from other areas of the Empire remain more difficult to interpret, as much more numerous hoards require a great deal of scrutiny in order to avoid our being misled by insufficient specific information on individual hoards. At the same time, the reign of Magnentius introduces additional causes of discontinuity in the West. Nonetheless, 353-354 seems a significant turning point in Gaul, Germany, Britain, but less so in provinces with no active mints like Iberia or Africa, and the proportion of AE2 falls after 354 generally speaking, although it is hard to distinguish between the effects of the monetary reform and the possible demonetization of Magnentius’s coinage.40 It has been suggested as well that 348 represented a major date in Britain’s hoard structure, although several of the hoards in support of this scenario fail that test.41 In any case, we should not expect perfect discontinuity, as premodern coinage reforms often achieved only a rather low degree of efficiency.42

2.3 Analyzing the potential issues behind the reform

What could have been the rationale behind a reform aiming at replacing the old coinage with new and multiplying all prices about ten-fold – or, to put matters more precisely, increasing massively the notional purchasing power of the new version of the most common billon coin?

Most currency changes tend to comply with a limited set of conditions:

The early 350s situation fits well with the first condition. At that point of time, a multiplicity of seemingly very comparable AE3 nummi issued during the 330s and 340s, incorporating varying and significantly different proportions of silver, circulated together alongside more recent AE2s belonging to at least two standards, necessarily harming the smooth functioning of the coinage system. The imperial authorities must had been concerned for a while by the confused state of the billon coinage pool, as shown by the 341-345 period, when it seems no billon coinage was minted at all,43 and again in 348-352 with the reintroduction of a heavy AE2 with higher silver content. In such a context, a massive recall of all previous coins must have been considered for some time in order to address the level of confusion and heterogeneity into which the system had fallen. Maybe the imperial authorities had stopped short of such a measure previously, because of its huge logistical challenge.

The reform of 348 might have aimed at some form of deflation, in a manner reminiscent of 324 under Licinius. A hint of such official concern may be found in the Codes. The 349 C. Th. IX, 21, 6 edict forbids the extraction of silver from the pecunia maiorina (the large AE2), implying that the coin’s commodity value stood higher than its notional value, possibly linked to a relative scarcity of silver, and pointing possibly to a relative lack of silver supply. In other words, it was most likely undervalued, a rare situation for the billon currency, explaining why it became an easy target for forgers.44 The most rational hypothesis is that the authorities had made an attempt at lowering current prices by issuing higher quality coins at relatively low face values.

Based on this assumption, one may try to determine the early large AE2’s most likely value in units of account. In 338-341, gold was worth 13,200 talents per pound, while 348-351 prices of wheat are about two-fold their 338-341 level.45 This leads to about 26,000 talents per pound of gold in 348. Using the gold: silver: bronze 1: 12: 1,440 respective ratios from 301 would value the commodity price of the large AE2 at about 1,500 denarii. If copper had lost some relative value by that time,46 the price would be closer to 1,300 denarii. As the coin was undervalued, a 1,000 or 1,250 denarii value would fit quite neatly. The same computations applied to the post-348 AE3 leads to a 320-360 denarii range, leading us to suppose a 250 rather than 100 denarii face value,47 although it did not prevent the continuous use of the term centenionalis to designate the coin.48  

This attempt at stabilizing the price level failed for range of reasons – public expectations of higher prices in future since inflationary mechanisms had taken over, a fundamental lack of trust that led to the withdrawal, hoarding or melting of the relatively high commodity value AE2 coins (hence C. Th. IX, 21, 6 in 349), the inability of the authorities to issue such a high value coinage in sufficient quantities, and finally Magnentius’s usurpation in 350, leading to monetary competition between several authorities. This would explain the rapid debasement of the AE2, ensuring that the State would stop losing money in minting it, but leading to more monetary confusion as a result. All the conditions for a major currency reform were gathered

As we have seen, at some point in or before 354 many former coins were recalled or removed from circulation, while the new AE3 Falling Horseman was introduced.49 Even with a coinage exchange to address the confusing state of the pool of currency in circulation, and new overvalued coins ensuring some price increase, nothing in this mechanism implies the massive price jump observed around 351-353. Pricing the new AE3 with its better silver content (about 0.5-0.6% instead of 0.2%) at 1,000 denarii would probably have been enough to ensure an acceptable overvaluation with respect to its commodity value and would have triggered a modest price increase in return, possibly of a two-fold order of magnitude. As a coin worth at least a myriad (10,000 denarii) is attested by then and the new AE3 stands as the only possible candidate,50 this implies some fateful and unexpected decision by the authorities, breaking with previous policies that had ensured rough correspondence between commodity and currency values for several decades – with the exception of Licinius’s 12.5 denarii coinage. The degree of overvaluation of the new AE3 was so high initially that ‘many Constantinian issues are overstruck by false fel. temp. reparatio dies in and after 354’.51

The reasons that triggered such a discontinuous step remain extremely elusive. Would the imperial treasury have benefited much from a sudden increase in denarii-denominated notional prices, because of significant expenses denominated in base coinage? We do not think so. By that time, compulsory purchases of gold and silver were long past, and there was little benefit to expect on that side. The tariffing in gold carats or fractions of solidi of most money taxes had rendered requisitions obsolete. In P. Oxy. 16.1905, dating either from the mid-350s or the early 370s,52 the four levies denominated in denarii represented a small proportion of the entire tax burden, which was essentially levied in gold and agricultural goods.53 Billon coinage could not have represented a major component of the overall imperial income by the early 350s. At the same time, the army’s regular pay in billon had by then collapsed to an infinitesimal value compared to gratifications and enlistment-linked bonuses, all denominated in gold or silver.54 Papyri and imperial regulations testify as well to a decisive shift from billon towards gold and silver from around the turn of the century onward (on the papyri, see Part 1 of this paper).

This shift from commodities or billon to gold does not occur at the same time for each single contribution and does not necessarily mean the complete disappearance of base coinage in tax settlements, as is shown by such an imperial decision as late as C. Th. XI, 28, 9 (414), but the overall trend is very clear.55 For instance, billon coinage tariffs are known for the components of the vestis militaris until the early 340s. C. Th. XI, 9, 2 (337), X, 8, 4 (346), IX, 42, 7 (369) imply it was not settled in precious metal necessarily, while C. Th. VII, 6, 4 (396) stipulates that a vestis is worth 1 solidus instead of 2 tremisses, and the late fourth-century P. Ross. Georg. 5.61 provides gold valuations – although actual clothing equipment must have been possible yet as per C. Th. X, 9, 2 (395). The supplies provided to the recruits or primipilum is settled in billon in 326/7 (P. Mich. Inv. 1378) and prior to this date (P. Cair. Isid. 53, 59, 60 and 61) but in gold in P. Oxy. 16.1905 and P. Lips. 87 (second half of the fourth century). Similarly, the comparatio mularum vel equorum, does not belong to the gold and silver tax receivables in C. Th. XIII, 3, 2 (320, 326 or 354), but appears in gold in P. Oxy. 16.1905 and is so tariffed by imperial regulations dated 367 and 401 (C. Th. XI, 17, 1-2), with a similar trend for the aurum tironicum. This does not mean the average taxpayer never used base metal coinage to settle his tax liabilities, but that the personnel involved in the taxation process had to remit gold coins to the imperial treasury generally speaking.56 In private transactions, the rise of the solidus is impressive on each side of a 350 divide (see above, section 1). In the American Numismatic Society collection, 145 gold coins of all types belong to the 294-350 period and 330 to the next 58 years which, unless hoarding would have been significantly higher during the latter part of the fourth century, points toward a higher output. This is confirmed in Gaul by a more than twice higher annual density for stray finds of gold coins after 364 compared to the earlier 284-363 period.57 Finally, the distribution of imperial decisions tariffed in solidi in the C. Th. and C. J. adds to a consistent overall picture. In the 16 books of the Codex Theodosianus for instance, the term solidus appears in 54 decisions, only 7 of them dating prior to 350. The 360-370 decade seems the turning point within the legal corpus, as 10 edicts for that sole period, i.e., more than during the entire 313-360 range. Adding the mentions of gold, silver, tremisses and semisses further confirms that skewed distribution.

Chart 3. Number of decisions in the Codex Theodosianus using the word solidus by decades

2.4 One possible rationale

What follows is highly hypothetical, is not based on any explicit ancient source, and aims solely at providing a scenario that ticks all the boxes of what we are observing. The State could not benefit in a very significant manner from a massive overvaluation of the billon coinage and/or a major manipulation of the monetary unit of account in a mostly commodity-based budgetary system (coined or uncoined gold and silver, grain, other staples) where the billon coinage’s role had become marginal in monetary value as far as the imperial budget was concerned. The sole reason behind such a discontinuous and massive change in the denarius-based price structure must then lie with an attempt at collecting the actual silver bullion contained in the billon coinage circulating all over the Empire.

We need to go back to the Codes at this point. The 354 or 356 C. Th. IX, 23, 1 edict in itself does not represent the demonetization decision. It simply reminds everyone that a wide range of old coins was already forbidden. The demonetization had occurred at some point prior to its date. As the State could not simply decree that old coins were worthless – especially following the recent 348 reform, it had to offer some form of exchange ratio between all legacy coins and the new issues, based on a comprehensive tariff. Unfortunately, the edict(s) that regulated the process is (are) not preserved by the Codes or by any papyrus.

In order to induce the holders of coins to bring their coins in for exchange, some incentive had to be offered to them. In late medieval or early modern Italian cities, 15th century England or 17th and 18th century France, for instance, successful coinage reforms normally offered such incentive mechanisms, whether new debased coins incorporated a higher face value than the old coins, or old coins were revalued by the State at a notional value higher than their last official price, before being demonetized and new coins with a lesser metallic value per currency unit issued so that the kingdom would benefit in the end and capture some precious metal through the coinage exchange. In essence, the State had to share its notional profit with the coins’ holders, otherwise the recall would be largely unsuccessful.58

By 351-353, the eastern monetary pool was constituted by the Felix Temporum Reparatio series from 348, including the surviving large AE2, the debased small AE2 (ca. 4.48 g, with at best 1% of silver), most of them of the Falling Horseman type in the East, and the AE3 at 2.68 g/0.20% silver, supplemented by a wide array of older Constantinian AE3s, with even more ancient base metal issues. As the early large and highly silvered AE2 Falling Horseman had been most likely priced originally at 1,250 denarii and undervalued, while the same mark of value A was kept on the later, debased issues of the small AE2, we will assume that the poorer AE2 adopted its notional value, with the AE3 still at 250 denarii.59 The remaining early large AE2s would have been revalued or at least valued by the market at 2,500 denarii, as it incorporated about twice the commodity content of its successor within the AE2 range.

Let us assume hypothetically that 10,000 denarii (a myriad) had been offered by the imperial authorities in return for each large, early AE2, being brought to the mint after it was demonetized, a figure significantly higher than its market value. At the same time, the smaller AE2 and AE3 denominations could have been priced proportionally to their relative commodity values, at 5,000 and 1,000 denarii respectively. During the coinage recall, the authorities would have returned to the exchange participants the new AE3 Falling Horseman, tariffed at 10,000 denarii as well, while demonetizing all the coins being recalled.

The large AE2’s commodity value (150 mg of silver + ca. 5 g of bronze) stood about five times higher than that of the new Falling Horseman AE3 (33 mg of silver + ca. 2 g of bronze). Such a mechanism would have secured a gross profit standing at 80% of the commodity value of the larger coin brought to the State – and a similar proportion of the silver included in all demonetized coins, provided that their conversion rates had stood proportional to their relative commodity content. Even though such considerations only allow for ball park estimates, as billon coinage’s effective silver content are notoriously hard to assess and fluctuated in any case, they may provide the best rationale behind the richer coins’ demonetization.

The holders of coins would have benefitted initially from the gap between the old and new notional prices offered for their existing coins, even though they had lost in commodity value and would ultimately be fooled, as prices would finally have to adjust at some point to the collapse in the commodity value of the reference coinage. The new price had to stand high enough, the threat of a looming demonetization adding some pressure to ensure that few would be inclined to evade. Yet in 354 or 356 C. Th. IX, 23, 1 had to remind everyone that old coins were forbidden, showing that some users had not fallen into the trap. As the new coins were initially massively overvalued, this would explain the overstriking with later dies of Constantinian issues irrespective of their higher silver content.60

In any case, the AE2 coins that had not been brought to the exchange would have been withdrawn by their owners from the monetary pool as their commodity value surged after the coinage change, hence their rarity from post-354 hoards.

If the exchange had involved valuing the large AE2 at 10,000 denarii prior to its removal, prices would have initially needed to rise about four-fold to compensate for the 2,500 to 10,000 denarii increase of the AE2 coin’s notional value. Then a further and more progressive five-fold adjustment would have been necessary to account for the loss of commodity value between the large AE2 and the new AE3 that replaced it. The combined price increase would have reached 20-fold eventually, an order of magnitude quite compatible with the papyrological documentation.61

Such a scenario would have allowed the State to profit from the coinage exchange without having recourse to excessive and ineffective coercion. It would explain the exclusive use of the Falling Horseman iconography by the new AE3 – because it had been mostly used for the highest of the post-348 denominations by Constantius II’s mints, highlighting that the new coins shared the same notional value as the large AE2s. Not so paradoxically, the State then would have implemented with success the very same metal extraction it had forbidden its citizens to undertake in 349 (C. Th. IX, 21, 6).

Prices did not adjust immediately. Some degree of monetary illusion – what economists call price stickiness - had to persist for some time and, as a result, AE3s remained significantly overvalued for a period of time; otherwise there would have been no incentive to produce all those Falling Horseman forgeries out of earlier Constantinian pieces. The undated P. Oxy. 34.2729 had rightly caught the attention of Roger Bagnall, Jean-Michel Carrié, Jean-Pierre Callu and Jean–Noël Barrandon since its 1968 edition by John Rea.62 It provides a solidus priced at 730 myriads, equivalent to a pound of gold worth 350,400 talents. If compared to a 15,000-20,000 talents range in the early 340s and the 648,000 and 969,200 talents recorded for the 350s and 360s respectively, it implies a dating quite soon after the 351-353 price increase.63 The papyrus may be one of the closest pieces of evidence we possess with respect to the coinage change, alongside SB 16.12828, where a pound of gold is priced at 276,480 talents. P. Oxy. 34.2729 provides us with another very interesting item of price information: a copper pound is worth 18 argyra, the coin being necessarily the post-reform Falling Horseman AE3.64 Its equivalent bronze value was about 4.60 g based on its silver content and a 1:120 ratio between silver and copper, and 18 of these coins would bring about 83 g, or 25% of the weight of  copper (1 lb.) they are deemed to purchase, provided copper and bronze commanded approximately similar values. Does it mean it was overvalued four times at that point? The issue here is that several types of bronze or copper could command different prices. In 301, the Price Edict had listed a ‘Cyprian bronze’ or ‘copper’ at 75 denarii per pound, while two forms of bronze were priced at 50 and 60 denarii respectively. Values of copper and bronze obviously depended on their relative quality and whether or not the metal was in a cast or a worked form.65 In fact, an implicit gold to copper commodity ratio is computable, as a gold solidus of 1/72 of a pound is worth 730 myriads and a pound of copper is worth 18 argyra – each argyron being a myriad of 10,000 denarii: 730 x 72 / 18 = 2,920. This fits with the range of ratios computed by Bagnall for cast bronze in the fourth century: 3,300; 1,935; a 1,987-2,635 range; 2,520.66 Assuming a ca. 1:14 ratio between gold and silver, the implicit silver to copper ratio would have stood actually closer to 1:200 than the Price Edict 1:120. Assuming a rough identity between prices of raw copper and bronze, this would imply that the new Falling Horseman AE3 (ca. 33 mg of silver + ca. 2 g of bronze) notional value was about twice its commodity value at the time of P. Oxy. 34.2729, since 18 of them represented about 156 g of bronze equivalent and purchased at that time one Roman pound of copper, i.e., ca. 323 g. This represents an extremely interesting observation, as it stands within the range that had been estimated by J. –P. Callu for the 352-361 period.67

The papyrus would testify to the period of time during which prices adjusted to the coinage change, hence the particular anxiety displayed by its writer about securing the purchase of certain list of objects as quickly as possible, as prices moved rapidly. After the adjustment was over, the AE3 would have been valued following its commodity content, preventing the state from further accruing more seignoriage while minting it. As long as billon coins retained their overvaluation, the public would have tended to hoard gold and silver coins. This may explain the restriction imposed upon the transportation of billon coins in 354, in order to force precious coins’ holders to put them back into circulation.68 Currency units used by fourth-century loan contracts confirm the very marginal use of gold and silver until that period, as monetary loans from 300-363 use myriads and talents of denarii.69

What about the exchange mechanism itself? In 18th century France, existing coins had to be brought to the mint, where new coins were issued, but normally the new coins were not immediately provided in return – unless they were simply overstruck from existing coins with some distinctive mark when the exchange did not change the coins’ physical characteristics but simply their values in units of accounts. Coin holders were credited with a claim against the State, materialized after 1700 by interest-bearing billets de monoye that would later be used to obtain the new coins. Such a delayed process ensured that the State did not have to pile up stocks of metal ahead of the recall and could recycle the metallic content of the recalled coins. However, no such mechanisms existed in Antiquity. In contrast, the frequent coin-recalls in medieval Europe led to an immediate exchange, usually at the time of major local fairs, implying the previous purchase or the availability of metallic resources.70 But these demonetizations involved quite limited amounts of coinage overall. It seems doubtful imperial mints held such large metallic supplies that they could have handled a general recoinage of a much larger scale without having hold of a proportion of the existing coinage first. Fiscal mechanisms could have helped, as the state did receive a proportion of the coins it had minted through tax payment. However, it would have taken years to achieve a substantial replacement of the coins’ supply by this sole channel, as all base coinage taxes must have represented quite a small proportion of the entire coinage supply.  

In fact, the likely geographically sequential implementation of the reform between 351 and 354 would have decreased the need for such metallic supplies Empire-wide, while a significant proportion of coins’ holders practiced hoarding of high quality coins and further reduced the pressure on the imperial authorities. Finally, are we able exclude with absolute certainty the possibility that, in some cases, coins’ holders were initially credited with a corresponding balance in denarii in some form of accounting maintained by tax officials acting as intermediaries and were provided with the physical coins at a later stage? Whatever the mechanisms might have been, the progressive disappearance of the 348 reform coinage and its replacement on or before 354 led to a significant withdrawal of silver from the base coinage pool.

2.5 A limited but substantial reinjection of silver into the Roman economy

Gold becomes pervasive after the 350s, probably owing to new sources of supply, as platinum content increases dramatically in the gold used by the coins of the period.71 Interestingly, silver coinage is produced in copious issues in the late 350s and early 360s, compared to the 330s and 340s.72 (Plate 15). Even more striking, recent metallic analyses have concluded that the same type of silver ore had been used for the production of billon coins between 294 and 364 and silver coins between 364 and 375.73 As post-364 base metal coins do not display any sign of voluntarily added silver, this analysis implies that the silver used to produce billon coins between Diocletian and Julian could have been recycled into the minting of Valentinianic silver coins. No analysis of silver coins produced prior to 364 was performed, but since no other coinage reform associated with a discontinuity in the pricing system occurred after the 351-353 gap, there is no reason to speculate that silver coins minted under Constantius II and Julian would not share the same metallic content with those produced under Valentinian. This lends further credence to the most likely motivation behind the 351-353 coinage reform – extracting as much silver as possible from the billon coinage pool. Imposing significant losses of silver commodity on all holders of billon coins through a compulsory currency exchange becomes the most likely explanation behind such greater availability of silver, unless one would hypothesize the existence of a single cohesive source of fresh silver throughout the century. Such a scenario is contradicted by the visible scarcity of silver within the coinage pool from Diocletian until Constantius II, as silver used for the production of billon coins was recycled from one series to the next, 74 and silver coinage remained rare after 301. In simple terms, provided all billon coins had been recalled successfully, the authorities would have acquired 80% of the silver that had circulated throughout the Empire incorporated into the earlier billon coins.

We have no way to evaluate what such a mass represented, but it must have been substantial. In rough orders of magnitude, if billon coins had weighed about 40% of the overall monetary supply in value during the first half of the fourth century,75 and if they had contained silver that represented roughly an aggregate 60% of that coinage’s value between the AE2s and AE3s,76 assuming that 75% of the older coins were brought to the exchange and 80% of their commodity removed from the coinage, the State would have acquired a quantity of silver worth about 14% of the entire monetary supply in value. Provided coinage supply and ‘Gross Domestic Production’ had stood in the same rough order of magnitude, and that the overall tax extraction rate did not exceed 10%, the State would have accumulated a quantity of silver worth almost one and a half times its annual budget. We are not trying to offer a quantitative estimate of how much silver would have been captured, as there is no available estimate of the size of the billon issues, but simply to suggest an order of magnitude – which would have been very high.

Hendy’s estimate of the Roman Empire’s budget under Justinian reaches 5 to 6 million solidi.77 If the Empire had earned about 50% more under Constantius II, its income would then have amounted to the equivalent of about 8-9 million solidi, and the quantity of silver gathered through the exchange close to the value of 12 million solidi or about 167,000 pounds of gold-equivalent, about 2,000,000 pounds of silver. Such orders of magnitude could be compared to the quantity of gold and silver invested in the botched attempt at recovering North Africa in 468, said to be 64,000 pounds of gold and 700,000 pounds of silver,78 a good indication of the significance of the one-shot silver extraction that would have been achieved in or by 353. In 359 Julian promised his army a gratification of five solidi and one pound of silver per soldier. This would have amounted to about 25,000 pounds of silver, based on the size of his Gallic army. Similarly, Julian claims in one of his letters that Florentius, Constantius II’s praetorian prefect, had been willing to offer 2,000 pounds of silver to the barbarian invaders of Gaul rather than engage them in battle.79 Such reappearance of raw silver in our sources would have been facilitated by the proceeds of the coin recall in the northwestern provinces.

The similarity with the imperial compulsory purchases of gold and silver that had taken place from Diocletian until the early 330s is striking: in both cases, the Empire aimed at restoring a fully functional precious metal coinage. But as before, it represented a one-time measure that could not impact fundamentally the Empire’s long term overall financial capacities or be repeated. Nevertheless, it must have represented a substantial lump sum at that time, just as the increased availability of gold was about to address imperial financial imbalances in a more fundamental manner. The size of the one-shot influx, combined with a lack of sustainable sources of silver, could be correlated with the silver coinage reform initiated by Constantius II in ca. 355 with a reduction in the coinage’s weight but a massive increase in minting volumes.80 A sudden influx of new silver metal is highly consistent with the actual silver coinage minting activity, provided surviving coins reflect it.

Chart 4: Number of coins from the ANS collection, sorted by minting year, 296-525 CE.
Chart 5: Number of coins from the ANS collection, sorted by issuer, 296-525 CE.

Studying fourth century coins minted in Gaul, Depeyrot ends with a very similar profile, with 2,560 silver coins minted during the 358-364 period vs. 29 only between 348 and 357 among his sample! The next period, 364-402, witnesses a progressive decline in silver issuances after a peak around 367-378. Trier, the main mint by far for silver, produced 2,774 silver coins between 364 and 402. At least 2,728 of them are dated before 378.81 The 945 coins minted in Milan from 383 until 408 after the closure of Trier and found in Gaul and Britain do not fundamentally alter a general pattern of progressive disappearance of silver by the end of the fourth century – especially as weight standards were reduced.82 At the same time, base metal coins minted after 364 do not display traces of intentionally added silver any more.83 Overall, the 358-378 period stands out in a very significant manner.

The textual record of legal practice displays similar patterns, with silver fines being concentrated essentially in a 335-380 time range, while gold fines appear mostly after 355, displaying a clear chronological shift compared to silver.84 This is confirmed by the overall frequency of the use of argentum in the Codex Theodosianus.

Chart 6: Frequency of the use of the word argentum in the Codex Theodosianus.

The simultaneity between the surge in the supply of gold and silver coinages, the monetary change of 351/3, and the durability of the gold revolution leads to some intriguing questions – did the actual discovery of very significant new mining sources of gold lead the authorities to seek more silver as well in an attempt at resurrecting the trimetallic imperial system of the first two centuries, leading to this effective requisition of silver from the billon coinage pool? The promotion of precious metal coinage as testified to by the 354 or 356 edict would have failed initially, as gold and silver would have been hoarded as a result of the new billon coinage’s excessive devaluation. Further and sustained availability of gold was needed to finally support the imperial policy. In essence, the imperial authorities succeeded only when a consistent supply of new precious metal existed. This materialized only for gold in the long run, hence the essentially bimetallic gold and copper nature of the fifth- and sixth-century coinage. (Plates 16 and 17).

2.6  Conclusion

The suggested process is not directly attested by any ancient document or testimony, and other scenarios could be suggested. Nevertheless, the early fel. temp. reparatio coinage’s replacement by a not-so-different coinage while all prices rose about twenty-fold in a relatively short period of time implies a manipulation of the basic coin’s notional value. The AE3 Falling Horseman must have been valued much higher in currency units than its heavier and finer AE2 predecessors had been for such a price jump to occur. Since the imperial budget did not mainly rely on billon coinage by the second half of the fourth century, the main reason for the coin recall must have been commodity-driven. In other words, the authorities aimed at increasing their sources of silver. One way or another, a coin conversion had to take place, consistent with the discontinuity displayed by the hoards around the 348-354 divide. Whatever process had taken place cannot escape this set of constraints, effectively reducing the range of possible scenarios.

This currency change was essentially a one-shot possibility: once everyone had grasped the full extent of the price increase, fiduciarity acceptance mostly disappeared. This explains why a later fourth century edict values solidi against bronze bullion directly, which must have implied a delicate balance between weighing and counting actual coins.85 By that time, bronze coins must have mostly retained their intrinsic value as the public had lost trust into the non-commodity component of its currency system. At that point, a coin was worth whatever metal it incorporated plus minting costs, no less, no more. When heavy base metal coins were reintroduced into the Mediterranean monetary system during the later part of the fifth century, they used plain bronze and avoided billon, narrowing the range of manipulations by minting authorities that billon had allowed.


The references for Roman Imperial coinage types adapt the format defined in, derived from the standard reference work authority of the Roman Empire (The Roman Imperial Coinage volumes published by Spinks) and available in Online Coins of the Roman Empire ( . For instance, represents the type 1a minted under Augustus, as published by The Roman Imperial Coinage, Volume I, Revised Edition. Spinks has published to date ten volumes of RIC, of which two were later revised. From RIC volumes 6 to 9, mints instead of emperors are used as the main way to reference coin types. The photographs, all available in the American Numismatic Society opensoucre database Mantis (, are shown at the same size and so do not depict the coins to scale.

The date or date range given for each coin reflect those published by the editors of the corresponding RIC volumes. Our paper may impact some of those datings to a limited extent, especially around the critical period of 351-354.