Growth and change? Roman Economic History and the archaeological record

8-06-2012, 14:00

The key story in Roman Economic History is, perhaps inevitably, the combination of apparently unprecedented levels of economic development and integration with a subsequent and apparently rather radical decline. The two big issues behind that story are, basically, where the development comes from, and where it goes. As Saller and, more recently, Scheidel have pointed out quite rightly, it is, in discussing these issues, no longer enough to just sum up evidence for unprecedented extraction, trade or production – few, at this point, would, I feel, disagree with the general idea that the Mediterranean and European economies of the late Hellenistic and Early Imperial periods easily surpassed anything that had preceded them in this part of the world. The real questions at stake are whether this was the result from extensive or intensive growth, and whether the growth eventually was terminated by, to use Scheidel’s words, ‘endogenous’ or ‘exogenous’ factors. The moment we start considering these questions, we are in contested territory, in the dark, and in the middle, surrounded by, on our left side, neo-primitivists claiming that the Roman Economy performed, per head of the population, only very marginally better than the Hellenistic economy, and that growth stopped because the Empire hit the Malthusian ceiling. On the right are the neo-modernists telling us that Roman growth was at least partially intensive, and that it was brought to a halt by plague, invasions and political turmoil – developments that the other side might easily set away as symptoms, rather than causes, of the decline. Neither side has, as we speak, found the key argument to settle the issue for the foreseeable future, and few of the datasets commonly used in the economy debate seem to be able to do so.

Ostia, V vii 3, a: Rinsing Complex - View from the place where the water-supply mouthed. Clothes went into the opposite direction. (Photo: Miko Flohr [2006])

1. Introduction

This paper is not going to solve this issue definitively. What it is going to do is thinking, or perhaps better rethinking, from the perspective of what we have done, thus far, at the Roman Economy Project in Oxford, what archaeological evidence may have to contribute to this debate. Our project, as some of you may know, started back in 2005 with the idea that it might be useful to explore the possibilities of large, quantifiable datasets for understanding the extent and nature of Roman growth. This has resulted in a large quantity of raw data produced by the project’s research assistants, a series of conferences and conference proceedings, and a list of publications by project members. As far as the archaeological evidence is concerned, particular energy has been invested in improving our understanding of the material record of shipwrecks, mines, fish-processing industries, wine- and olive presses, and evidence for the application of water-power.

These are all complex, multi-faceted datasets that do not easily allow for generalizations, and the idea was not, of course, that we would gather the data, press the red button, and instantly receive full understanding of Roman Economic history. Rather, the idea was that one of the ways to get a reliable idea of what the evidence for production and trade really means is to analyze it at the macro-level. The project, thus, has, from the start, taken a strongly inductive methodological position, and I think that that is a position the debate also rather urgently needs – bridging the wide gap between the many macro-scale deductive approaches that dominate the field among hardcore Roman economic historians and the small-scale, often only locally oriented inductive approaches that tend to prevail among Roman field archaeologists.

In this paper, I am going to focus on one key theme in Roman Economic History – production – and discuss this in the light of the data sets studied by the project, before coming back on the issue of Roman economic growth and on where this debate might or might not go next. Within the theme of ‘production’ I will particularly focus on two variables that play a key role in the discussion – scale and technology – and I will emphasize how an approach that uses the quantifiable elements in the evidence can help us qualify differentiation and understand what is going on – even without producing graphs that can be combined into a fever chart mapping Roman Economic growth and decline through the centuries. This is not necessarily something that is to be missed: I have always thought that we do the complexity of our data more justice if we do not hide ourselves behind the false certainty of a line that goes up and down.

2. Scale and intensification

As far as scale is concerned, our project has developed a keen interest in manufacturing and food processing. I am going to focus on three data-sets here – one related to the textile economy, and two others related to the food economy. Of course, our current archaeological record is highly selective and biased, and there is no way you can start to speculate about the amounts of information that still await discovery or have gone lost unrecorded, so it would be pointless just to assemble the data, plot them on a map or timescale and read off the economic geography of the Roman world. In making sense of these data it is relevant to be able to make the internal discrepancies and developments within these data-sets visible.

A key issue is the chronological and geographical spread of differences in scale. Scale, of course, is a product of investment, and large-scale investment is positively related to expectations about large-scale demand: there is little point in financing or building a large-scale production facility unless you expect at least some return on investment. At the same time, the scale of production facilities is measurable through their material remains – not in an absolute sense – it doesn’t give clues about absolute output – but in a relative sense: material remains usually allow us to establish the difference in maximum production capacity of workshops.

By quantifying and comparing these production capacities in larger numbers, it becomes possible to build up a contextualized picture of varying or even changing investment strategies in the Roman world – and one that goes beyond the simplistic and anecdotal ‘hey, this is a big workshop, which shows that everything became bigger in the Imperial period’. In general, one can say that the preindustrial small scale remained the norm throughout the Roman World, but that the significant development is the emergence of larger scale establishments in certain places, in certain periods. The question is where, when, and why, and what the implications for Roman Economic History should be.

2.1. Fulling

One data-set that can be analyzed in this way is the evidence for fulling in the Roman world. Fulling is a process in which newly woven and used clothes are chemically degreased and then polished to enhance their visual appearance and make them more comfortable to wear. In the first phase of this process, clothes were treated with chemicals by trampling them under the feet in specialized fulling installations that consisted of a tub embedded in the floor surrounded by low walls that could serve as support for workers. As each installation provided place for one worker, the number of these fulling installations gives an indication of the maximum capacity of each workshop.

: Capacity of fulling workshops - (Photo: Miko Flohr [2010])

As you can see on this graph, we have a relatively limited dataset of some 24 workshops of which we know the capacity. They are all from Italy, with one exception, and the overwhelming majority of all workshops is very small – having only two or three fulling stalls. There is a small middle group of medium-sized fullonicae from Pompeii, and then there are five workshops of exceptionally large size – a slightly smaller one in Florence, but the other four all closely related to the city of Rome itself: three in Ostia, and the largest one just east of the city of Rome close to the Via Praenestina. All date to the second century AD, and most remained in use until at least the third century.

So there is evidence for large scale investment in fulling in the high empire, but what does it economically mean? The clue lies with the establishments in Rome and Ostia: these clearly seem to serve the local metropolitan market and are situated at places where transport lines came together: fulling was not part of the primary production chain but performed as an additional service by clothing traders before they sold their wares on the Roman market. The size and stability of this market made it feasible to invest on the large scale and enjoy economies of scale by organizing the production process in larger workshops. This is likely to have reduced overhead costs: it is to be emphasized that these workshops are very thoroughly organized and seem to have employed a rather rigid division of labour.

A similar development is visible in the mills/bakeries of the Roman metropolitan area, albeit on a slightly smaller scale. These large workshops show the flexibility of the Roman economic system: while small-scale remained the norm, the specific market circumstances in the early imperial metropolis fostered the emergence of other forms of workshops on a much larger scale.

2.2. Olive-Oil and wine presses

Unfortunately, at this point, the fulling workshops are the only urban workshops which can be analyzed in this way, and the limited geographical extent of this dataset does not really help us much beyond Italy. Moreover, it is one thing to invest on a larger scale when there is a large and stable local market and information about it is easily accessible, and it is another thing to do so when the market you are aiming at is much further away. The other two production-related datasets I want to discuss are both geographically more widespread and less close to the market they were serving.

: Capacity of Roman farms with two or more presses - (Photo: Miko Flohr [2012])

In our database of olive-oil and wine-presses, we recorded published press facilities from the Roman world where more than one press was found. There are 229 sites in our database and as you can see, most of them are concentrated in the smaller part of the spectrum: three quarters are farmsteads with two or three presses, and only eleven percent – or 25 establishments – have more than five presses. Yet, at the same time, the largest three have 17, 18 and 21 presses respectively. Again, the differentiation within the dataset is substantial.  The press sites are hard to date precisely and more work on that is certainly necessary, but most evidence suggests an imperial horizon in the first two centuries of our era.

: Press sites by country - (Photo: Miko Flohr [2012])

Yet, highly relevant is the relation between differentiation in scale and the geographical spread of the data. This is best shown by comparing the geographical spread of the entire data set with that of the large-scale pressing sites with more than five presses each. The fact that the graphs look completely different immediately suggests that large-scale pressing facilities were concentrated in a limited number of regions. Note the complete absence of Italy from the right graph, and the strong prominence of especially Tunisia and Algeria, but also of Croatia: two of the establishments in this group are from Istria. If we divide this by continent, this is how it looks: 80% of the large-scale pressing establishments is situated in Africa.

: Press sites by continent - (Photo: Miko Flohr [2012])

This can hardly be due to a bias in our evidence. Generally, most of Africa has received much less attention from archaeologists, and the complete lack of such large-scale establishments in Europe except for two examples in Istria suggests that they were much rarer on this side of the Mediterranean, suggesting, in turn, different patterns in landholding and investment strategies, with a much larger role for elite power play.

The question is what these large-scale press factories achieved in terms of return on investment: there is, in itself, little efficiency gain in just putting together a lot of presses as you would need equally larger stocks of people to man them. So again, the gain seems to be in the overhead costs and in the logistics. By centralizing the pressing procedure, presses and manpower could be used more efficiently, and transport could be coordinated better. Maybe, the logistics of bottling also played a role.

The scale and ambition behind this investment also tell us something about the economic integration of the Roman world: while we don’t know where the produce of these factories went, we do know it was transported over a rather large distance, and the geography suggests that they were consumed outside Africa. So they were created by economic strategies operating on a supra-regional, if not empire-wide basis.

2.3. Fish-processing plants

The same is also true for the large-scale fish-processing plants found on the Mediterranean and Atlantic coastline. In these factories, fish was salted and processed into a variety of fish sauces. The quantity of fish that could be processed in each factory was, of course, positively related to the capacity of the large vats in which the fish was salted. Not of all known factories, the capacity is known, but of some factories, we know the sizes of their salting vats. Moreover, we also know relatively precise when these factories were constructed. Numerically, most date between 50 BC and 50 AD – of the 33 workshops of which we know the size, 14 date from this period.

: Capacity of fish salting establishments - [source]

If we rank the factories according to size, the picture is rather comparable to that of the fulleries: many are very small, but a couple stand out. As with the fulleries, and with the press factories, the benefit of the scale is in the overhead costs and logistics: these factories simply could do more with less people.

Geographically, there is a strong emphasis on the southern coast of the west Mediterranean with the exception of Algeria, and the straits of the Gibraltar. As Andrew Wilson and several other people have argued, this has to do with the migratory patterns of fish species and must be seen as a relatively unbiased picture of historical reality – the gap in Algeria being caused by the inaccessibility of its archaeological record. In any case, these factories catered for an empire-wide market. This has also been archaeologically attested: preserved fish from the Mediterranean has been found at several places in Roman Europe.

: Fish salting factories: construction - number of factories versus aggregate capacity [source]

Yet, even within this picture, the larger establishments are not equally divided over time and space. If we look at the moment at which these large factories were first constructed there is a strong emphasis on the period second century AD: of the eight workshops with a aggregate vat capacity of more than 100 cubic metres, five were constructed in this period. So if we compare the graph with the number of workshops constructed in each period with a graph showing the aggregate capacity constructed in each period, the balance clearly shifts to the late first and early second century AD. This, apparently, was a period in which investors happily invested on a large scale in fish salting capacities. However, interestingly, there also is a geographical divide here: as Andrew Wilson already noticed, the large factories that were constructed in the second half of the first century all come from the region around the straits of Gibraltar, whereas those constructed after 100 AD come from the Atlantic coast of Northern France, in Brittany. Perhaps, this reflects a development in the market situation on an empire-wide level. However, more data are needed to be able to fully understand this differentiation. 

2.4. Discussion

None of these datasets on its own is able to say an awful lot about the economic history of production in the Roman world, yet if their internal variation is analyzed as we have thus done before, and if they are then compared, it is obvious that the trends clearly point to the same direction – even if they cannot be visually expressed in a graph: in all these datasets, the scale of investment clearly peaks in the first two centuries AD. This increase in scale implies, even without technological change, an increase in efficiency as it saves on overhead and organizational costs. As you see this increase in scale in several places and in several industries, it seems that certain groups of Romans invested massively in reducing the organizational costs of processing foods and goods. This points to the existence of intensive forms of growth that can be better understood if more datasets like these are analyzed in this way – by making use of their internal differentiation in time and space. Moreover – it may be pointed out that such large-scale facilities are extremely rare before this period, which argues against the idea that economic growth in this period was just caused by the expansion of the Mediterranean technological system over Europe, as has been argued by some: the system not only expanded, it also was very much able to adapt to the new market circumstances created by political unification of the Roman world by creating new forms of organization. I would like to stress that archaeology is, much more than the historical record, able to show this differentiation and these changing strategies.

3. Technology

: archaeologically known water-mills dated before AD 700 - rate of increase in information [source]

The second issue I want to briefly focus on is technology, and particularly, our dramatically expanding insight into the application of water-power, which is now even radically different from when the project started in 2005 and when the below graph was made by Andrew Wilson in 2006 (working paper). The process currently seems to feed itself: we know more water mills and hence are also increasingly able to recognize the less well-preserved examples in the field. It is increasingly becoming clear that the application of water power was much more widespread than was assumed in the past, not only in the east, but also the west. Unpublished work by Jean-Pierre Brun presented at Oxford shows water-mills being widespread throughout Roman Gaul.

Also, it is now widely accepted that the emergence of water-driven mills dates back much earlier than originally was assumed, and must, again, be seen as, at least, originally, a development of the first centuries of our era. Again, it is worth pointing out that archaeology makes the difference in this respect: whereas literary references are basically restricted to authors from later antiquity, the emphasis in the archaeological material is strongly on the second and third centuries. If we would update the chart with discoveries and publications of the last five years, the balance would shift further towards this earlier period. Future discoveries are likely to further corroborate this picture. This is an area in which the project will do more work over the next year or so.

Most of the evidence for water-power is related to the food economy, and to the milling of grain. The economic impact of using water-power instead of animal or human power to mill grain seems straightforward as you need to spend less on labour costs or animal fodder, but its real impact has recently been doubted by Walter Scheidel (2009), based on three arguments.

  1. Milling was not a bottleneck in production, so an increase in milling capacity would not lead to an ‘economically meaningful outcome’.
  2. It is unlikely that Roman milling capacity was comparable to that of England in 1086, so the Roman economy was less advanced than a relatively marginal medieval economy.
  3. If there were any productivity gains from water-power, they might easily have been absorbed by population growth.

None of these arguments is really convincing. As to the first argument, as Wilson (2009) already argued in his response to Scheidel’s JRA article, in which he stated all this, milling probably was a bottleneck, as is suggested by the continuous improvements to milling technique throughout antiquity. One could add that even if it was not a real bottleneck, milling would have lowered production costs and would thus have freed up demand for other products, and the hands to make them. In this way, there is no fundamental difference between the effects of producing on a larger scale and that of applying water power. Scheidel’s second argument is irrelevant: we are not so much concerned with whether or not the Roman world was more or less advanced than Medieval England, but rather with how the Roman world itself developed economically. Even if the per capita mill-density of the Roman world in the high empire was much lower than that of eleventh century England, it can have presented a significant advance with respect to earlier periods.

This leaves us with Scheidel’s third argument, which essentially is Malthusian in nature: increased productivity leads to population growth which in turn lowers per capita productivity. Key problem with this argument is Scheidel’s presupposition that the emergence of water-power was related to increasing competition for land between humans and animals, which sounds attractive but is not supported by any evidence. It also seems to depart from the implicit assumption that economic strategies in the Roman world were shaped by necessity rather than by opportunity – you only change when you are forced by the circumstances, otherwise you’ll keep doing things the same way. I think that is too pessimistic a scenario: even with acknowledging the pre- or non-industrial constraints with which the Roman world had to deal, there is ample room for a scenario in which economic strategies were sometimes, or often, driven by opportunities. Yet even if Scheidel would be right, and the productivity gains of water-power would be eaten away by the extra mouths produced by its increased application, this still would imply, for the short term, per capita growth, and for the longer term, the potential to sustain more people from roughly the same material basis – which one technically cannot call per capita growth but certainly is a form of intensification. It belongs to the realm of intensive growth, not to that of extensive growth.

From this point of view there is thus little reason to be sceptical about what Roman historians or archaeologists can say about the implications of the uptake of water-power for our view on Roman economic history. The problems are much more pragmatic: the archaeological dataset currently is expanding at such a speed that our views might change rather dramatically over the next few years, and at the same time there clearly is much to be won in the eastern half of the empire, where the state of research is simply not at the same standard as it is in Roman Europe, so that much less is known about the application of water-power beyond some urban or large-scale examples. If it can be argued that per capita mill density in Roman Anatolia was as high as it seems to have been in Roman Gaul, we would simply have a rather strong case against those who would prefer to marginalize the impact of technological change on the Roman economy. Yet we need to work on the data.

4. Discussion

I would like to conclude with emphasizing two points. First, thorough study of large archaeological datasets allows us, basically, to get an understanding of what is going on in the Roman world in a way no other type of data will ever be able to do – maybe with the partial exception of epigraphy, but this, essentially, is  an archaeological dataset anyway. When it comes to production, material remains do not, of course, help us to quantify net output of workshops or to quantify trade flows – for that, the archaeological record is much too fragmentary and biased. Yet, studying the remains of production facilities in larger numbers does allow us to point to chronologically or geographically specific investment strategies that, in a more indirect way, tell us about the history of the Roman economy and its geographic emphases in a way simple observations in the field cannot do. Through quantified difference analysis, it becomes possible to qualify historical trends.

Secondly, the datasets studied within the Oxford Roman Economy Project are complex and biased and their interpretation is anything but straightforward. They may often be less easily quantifiable than we want them to be. Yet, they point to the same direction: the Roman economy not only became bigger, it also became increasingly efficient, and this increase in efficiency is not restricted to the period of conquest, but continues well into the second century AD, and, indeed, when one concerns water-power, much later as well. The Romans were, at least to some degree, very well able to adapt their economy to the changed circumstances of the empire and make use of the opportunities shaped by political unity. There was structural and intensive growth, and archaeology helps to understand how that shaped up.

One question that remains is of course to which extent this intensive growth is a phenomenon of the western half of the empire, and to which extent it can also be traced in the eastern half, where the Roman political unification may be thought to have had less impact on existing socioeconomic structures. Here, I think Scheidel rightly pointed to a gap in our knowledge and in current archaeological approaches, which predominantly have been focusing on the Latin-speaking half of the empire. The problem is a lack of data. For instance, much less is known about olive oil and wine pressing facilities. Can we find the large multiple-press installations also in Anatolia, or is the increase in maximum scale really restricted to Italy? If so, why? These are questions that need to be high on the agenda of archaeologists and Roman Economic historians. At this point, they simply cannot be answered.

Another question to be addressed concerns the Malthusian ceiling and the precise workings of the Malthusian mechanisms as far as the Roman world is concerned. Much has been made of the Medieval Black death, and its effects on labour costs and prices. As far as I know, it has not been universally accepted that this really was a Malthusian correction. The same goes for the Antonine plague. It is rather hard to understand to which extent an economic system is close to or hits a Malthusian ceiling, especially when you are talking about an empire of the size of the Roman Empire – maybe at some point, in some place or region within the Empire, a Malthusian check may have occurred, but I would argue that this can barely have been the case for the system as a whole – it would not only imply that all agriculturally useful land would have been used to its maximum capacity, for which there is simply no evidence at all, especially not after the second century AD, but also that all possibilities of intensification within the technological and economical system would have been exhausted. Paradoxically, anyone wishing to argue in favour of a Malthusian scenario would thus do best to join us in our search for large-scale investment and for the application of water-power.

This paper was presented at a study day at the University of Cologne on June 8, 2012, and incorporates work done by several members of the Oxford Roman Economy project, especially Andrew Wilson. Two of his working papers published on the web site of the Oxford Roman Economy Project discuss the evidence for fish salting and water mills to greater detail.


Bowman, A. and A. Wilson (2009). Quantifying the Roman Economy. Methods and Problems. Oxford Studies on the Roman Economy. Oxford.
Flohr, M. (2013). The World of the Fullo. Work, Economy and Society in Roman Italy. Oxford Studies on the Roman Economy. Oxford.
Scheidel, W. (2009). ‘In search of Roman economic growth’. JRA 22: 46–70.
Wilson, A. (2006). ‘Fishy Business: Roman exploitation of marine resources’. JRA 19: 525–537.
Wilson, A. (2009). ‘Indicators for Roman economic growth: a response to Walter Scheidel’. JRA 22: 71–82.
Wilson, A. (2009). ‘Approaches to Quantifying Roman Trade’, in A. Bowman and A. Wilson (eds), Quantifying the Roman Economy. Methods and Problems n/a. Oxford, 213–249.