• 1 This research was supported by the project “Average-Transaction Costs and Risk Management during t (...)

1This article introduces the Italian side of the AveTransRisk database and reflects generally on the process of creating a large online database of historical data in order to draw lessons for future scholars. The AveTransRisk database was recently published by the ERC project “Average-Transaction Costs and Risk Management during the First Globalization” (AveTransRisk), based at the University of Exeter (UK), and captures a wealth of data on early modern maritime trade, particularly in the Mediterranean basin.1 The article will discuss the particularities of the Italian data and the solutions put in place to capture it, outlining the choices that confronted the database creators during the four-year period of its construction. It will demonstrate the capabilities of the resulting database as a tool for historical research. It will provide a clear description of the documentation, the process of creation, the resulting product, and its potential uses, and conclude with some thoughts for the future. The article will first outline the context in which the Italian sources were produced, with particular attention on how the values contained within were arrived at. It will then outline the technical solutions adopted and show the database’s potential for historical analysis. The final section will reflect on the end result and the process of creation.

• 2 G. Felloni, 1998 [1978].

2The Italian data come from the port cities of Genoa and Tuscany, where the local state archives contain some of the best serial data available on historical general average (GA). AveTransRisk – based on the pioneering research of Giuseppe Felloni at the University of Genoa almost fifty years ago – thus adopts the Italian data as “standard” and the original database architecture was created in response to the way this data is structured.2 The article thus includes technical discussion of design decisions that were made with this context in mind. Nevertheless, although the two cities were only 150 km away as the ship sails, different political economies and administrative traditions impacted greatly upon the structure of the respective sources, something that the database endeavours to respect. AveTransRisk now contains information on 1,149 voyages whose documents are preserved in Genoa and 213 from Tuscany. More sources are still being uploaded as we write.

• 3 M. H. Sánchez et al., 2011.
• 4 G. Doria, 1995; C. Bitossi, 1990.

3Genoa is located at the Northern edge of the Tyrrhenian Gulf. During the early modern period, this port city was the capital of a small republic squeezed onto the Ligurian coast between Savoy to the West and Tuscany to the South. From 1528, it was a close ally of the Spanish Empire.3 The Genoese granted loans (the so-called asientos) to the Spanish monarchy and, in return, they obtained formal and informal privileges that allowed them to engage profitably in trade and finance operations on a global scale. The Republic’s independence and “free trade” also rested on this alliance. Genoa was an oligarchic Republic: the main patrician families, who often also owned fiefs inland and had multiple commercial and financial interests, shared amongst themselves all political power and controlled GA procedures. The ruling class coincided with the state’s economic elite.4

• 5 On Livorno see A. Prosperi, 2009; J.-P. Filippini, 1998; L. Frattarelli Fischer, 2018.
• 6 L. Frattarelli Fischer, 1993.
• 7 C. Tazzara, 2014.

4The Tuscan case, meanwhile, reflects a quite different political-economic set up. Most early modern maritime traffic in Tuscany arrived in the port of Livorno, a so-called “free port” which was established by the Medici Grand Dukes in the late-sixteenth century (though GA cases continued to be the prerogative of the court of the Consoli del Mare in nearby Pisa even after Livorno’s rise ended Pisa’s role as an international port).5 The conditions established in the free port granted persons of all religions and nationalities the right to trade freely, at least in theory. The free port also offered tax exemptions, eventually resulting in the abolition of all import and export duties in 1676.6 More generally, the institutional set up in Livorno gave significant leeway to merchants and adopted a general spirit of non-interference in their dealings.7 Livorno relied above all on settled communities of foreign merchants, especially those originating from North-Western Europe, and did not possess a strong native merchant corps of its own, in sharp contrast to Genoa where oligarchic nobles had historically controlled both the state and a significant segment of the maritime trade.

• 8 URL: http://humanities-research.exeter.ac.uk/avetransrisk/voyages/events/10027/, Voyage ID 10027.
• 9 Sea protests were declared in Genoa 677 times. An Average procedure could result in more than a se (...)

6Almost all early modern GA records, regardless of jurisdiction, rest upon two documentary pillars (see the companion piece in this issue for further details). The first is the narrative element: the “sea protest” or consolato in Italian. This was the shipmaster’s account of the voyage and the misfortunes that had occasioned the sacrifice of property or extraordinary expense. Sea protests could vary in their form and content but usually provide a rich array of quantifiable information. While some of this data directly concerns early modern maritime trade (ship types and tonnage, itineraries, voyage times), we also find abundant weather information (wind direction, storms, the presence of ice) which could be used for reconstructing historical climate patterns. For example, the sea protest drawn up for the voyage of the Speranza Incoronata in 1668/1669 from Arkhangelsk to Livorno tells us that the seasonal winter ice, which had blocked the exit of the harbour of Arkhangelsk, had finally melted to an extent sufficient for navigation on 21 May 1669.8 Finally, sea protests also contain social information, in part because they were witnessed by several members of the crew (giving us master’s and witnesses’ names, places of origin, and roles on board). This sea protest content does not tend to differ greatly between the two centres and could, at any rate, have been produced somewhere other than Genoa or Livorno as sea protests were normally written (“declared”) in the first port encountered after the accident. The table below lists the declaration places found so far in the Genoese Average procedures uploaded in AveTransRisk, with the exception of Genoa (Table 1).9

• 11 Before the civil statutes, the tribunal of the Rota Civile chose the experts for each Average case (...)

11The Genoese civil statutes drafted in 1589 ruled, among other things, on GA and the persons responsible for drawing up calculations.11 They were officially called calculators (calcolatori). They were selected by the Genoese Senate, remained in office for 18 months, signed all the calculations, and had their own specialised notary/chancellor with a renewable five-year mandate. The calculators listened to all parties involved (shipmaster, merchants, and any insurers) and their witnesses and then validated or nullified the sea protests presented. Following the report’s approval, calculators could order the unloading of the goods and mandate the presence of guards on the ship to prevent any fraud. During the seventeenth century, GA procedures in Genoa underwent only slight changes, which consisted in an increase in the responsibilities, competences, and control exercised by the maritime tribunal of the Conservatori del Mare, composed of members of the nobility. With the passing of time, the Conservatori gradually absorbed the calculators’ functions. By the 1660s they had complete control over the procedure and the calculators seem to disappear.

• 12 A. Addobbati & J. Dyble, 2021, p. 837.

12In Tuscany, the figures responsible for drawing up calculations changed over the period under consideration. The calculations up until the mid-seventeenth century were carried out by two merchants who were selected by lot: one had to be a “recognised” Florentine merchant, and the other a “recognised” Pisan merchant. Around the 1650s, this procedure appears to have fallen into disuse, and from this point on, two calculators were selected at the discretion of the Pisan Consoli del Mare themselves. Little is known about their identity or profession, though the Consoli drew on a wide range of candidates: in one case we can be sure that the calculator was a merchant based in Livorno but hailing from a family of Pisan lawyers.12 The layout of the calculations suggests that the calculators appear to have used the ship’s bill of lading to draw up the document.

• 13 A. Watson, 2011 [1998], vol. 2, p. 420.
• 14 See E. Maccioni, 2019; V. Piergiovanni, 2012; L. Tanzini, 2015.
• 15 G. M. Casaregi, 1802, p. 28.
• 16 J. Dyble, 2021, p. 156.
• 17 J. Dyble, 2021, p. 158.

13The Tuscan calculators were explicitly tasked with the job of valuation, though is not always clear on what basis this was carried out. The relevant normative material apprehends this problem in part, though this provides contradictory assertions on what practice should be adopted. The Lex Rhodia of the Corpus Iuris Civilis stated that contributing goods should be valued at the price they can fetch (i.e. the market price at the destination), while sacrificed goods should be valued at their purchase price “since what is made good is loss suffered not gain foregone”.13 The Llibre del Consolat de Mar, on the other hand, a highly influential collection of maritime customs issued in Barcelona during the fifteenth century,14 stated that cargo should be valued at the price of the port of origin (i.e. the purchase price) if lost in the first half of the voyage and at the destination (selling price) if lost in the second half.15 On an operational level, the only certainty is that a variety of valuation approaches were adopted and that there was no one fixed way to value the cargo. Sacrificed and saved cargo seems to have received the same value in most cases. Two cases from the Tuscan data clearly used the “rule of halves” approach outlined by the Llibre, but there is no indication of its use in several instances where we would expect to see it applied: in a few cases (with multiple jettisons of the same cargo during a voyage, for example) we can be sure that it was not.16 Very often, we simply cannot tell: the most likely occurrence is that all cargo was simply being valued at the Livornese market rate. A government report of a somewhat later date (1785) states that merchants submitted invoices of their goods in order to establish values, though it is not clear that this practice was being applied in our period, and we have never found any trace of these invoices among our surviving sources.17

• 18 R. Davis, 1957, p. 410. For further details see A. Addobbati & J. Dyble, 2021, p. 838.

14The Tuscan calculators were likewise responsible for judging the value of the ship. In one case, historical records preserved elsewhere give the age of the ship in the GA in question, and thus allow us to compare the value given by the Tuscans with Ralph Davis’ estimates about the value of seventeenth-century vessels over their working life. In this case, we know that the ship was at least 13 years old and probably older. According to Davis’ estimates it should have been worth no more than 4,700 pieces of eight, and no less than 2,300. The Tuscan estimate of 4,000 pieces of eight is perhaps a touch generous – we know that the ship was in a poor state of repair even before the vicissitudes of the voyage – but certainly not unreasonable.18

• 19 See S. Corrieri, 2005.
• 20 As an example of this shift in the contribution criteria see the two voyages IDs 51105 (total vess (...)
• 21 On the differences in the contributing criteria between Genoa and Barcelona see A. Iodice, 2023.
• 22 Archivio di Stato di Genova, Genoa, Conservatori del Mare 109, 1691. On Carlo Targa see M. G. Mere (...)

15Once the ship had been valued, the two centres proceeded differently regarding the ship’s contribution to the GA. The Llibre del Consolat suggested that the vessel should count for only half its value.19 The effect of this provision was to lessen the financial burden upon ship masters and/or owners, whose interest in the venture – and hence final contribution – was thus reduced: this was probably introduced to promote the health of the maritime transport sector. Yet the documents we found in Genoa and Pisa tell a different tale, a tale about the decision of Genoese and Livornese institutions to adapt the Consolat by tailoring it on their own needs. In fact it is only from the 1660s that we find this rule being applied in Genoa: the Genoese, reeling from the economic and social effects of the 1656 plague, tried to make the port more attractive to shipmasters by copying the practices current in nearby Livorno, where the Llibre del Consolat’s rule on the partial contribution of the ship was followed.20 Previously, the whole value of the vessel was taken into account, a move that favoured merchants. In order to capture this information accurately, the database thus had to include a field for the vessel’s contributing value, one for the vessel’s non-contributing value, and one for the vessel’s total value. In Tuscany, meanwhile, the ship always counted for only half its value. Freight likewise contributed for one-third of its value according to the Llibre del Consolat but contributed in its entirety in Genoese practice.21 The Genoese lawyer and jurist Carlo Targa explained, in a document hidden inside a folder of GA procedures, how the Genoese authorities chose not to follow the Llibre del Consolat in this respect.22 In Targa’s opinion, the value of the freights is part of the value of the cargo. Decreasing the contribution rate of the former without decreasing the contribution rate of the latter – as goods always contributed for their full value – would be unfair and would benefit the shipowners. In Tuscany, meanwhile, freights contributed for one third. This is the reason why even the freights’ tab is divided into three subsections.

• 23 J. Dyble, 2021, p. 163; see, for example, Archivio di Stato di Pisa, Consoli del Mare, Atti Civili (...)
• 24 On these dynamics, see J. Dyble, 2023, pp. 378-379.
• 25 A. Addobbati & J. Dyble, 2021; for more general diplomatic conditions between England and Tuscany s (...)
• 26 A. Addobbati & J. Dyble, 2021.

16Tuscany’s political economy likewise rendered the question of valuation complicated at times, specifically with regard to the valuation of lost or damaged ship’s equipment. Here Livorno’s reliance on Northern European traffic led to distortions by contemporary actors. The financial import of damages suffered by a ship was usually assessed by neutral experts nominated by the court, often naval carpenters attached to the galleys of the Order of St Stephen.23 From the second half of the seventeenth century, however, English and Dutch shipmasters were apparently allowed to submit their own unnotarised damage reports in which they told the court how much each lost or damaged part of the ship was worth. The total sum requested was then “reduced” by the Pisan court by an arbitrary amount, probably to give a semblance of probity.24 This practice reflected the commercial power held by these actors and the desire of the Tuscan authorities to keep them onside and to maintain a good working relationship with them at a time of diplomatic tension with their home states.25 Our research even turned up one verified case of fraud, in which the English shipmaster and the (mostly English) receiving merchants colluded to push through a spuriously large GA claim in favour of the master, whilst the master, for his part, promised to help the merchants recoup their larger contributions by defrauding the Tuscan customs house.26 Since this would have been presented to the Pisan Consoli del Mare as an amicable settlement, and since the English were very important to the port economy, the Consoli had every reason to wave it through, and there is some reason to think that this may not have been an isolated example.

• 27 A. Addobbati & J. Dyble, 2021, p. 837.

18In any case, and whatever the precise mode of valuation, it is worth bearing in mind that in the majority of cases, wildly inaccurate valuations could not be made to stick. The incentive of each interested party was to have their cargo valued at the lowest possible price: the lower the price, relative to the other property involved, the lower its owner’s contribution. Yet the fact that all interested parties faced the same incentive was itself a guard against falsely low valuation. Many receiving merchants would be receiving the same cargoes, making it difficult to favour particular parties. Moreover, an ever-present dynamic of the GA case is the tension between “hull” and “cargo”. The shipmaster would have hoped for a low valuation for his ship and freight, and a high valuation for the merchandise; the merchants hoped for the opposite. Save for cases of fraud like those outlined above, in which one party found a way to secretly “reimburse” the other for an unfavourable valuation, the presence of the shipmaster and at least some of the receiving merchants during a GA procedure meant that an acceptable balance had to be found between these opposing interests, and hence a mutually acceptable appraisal of goods, ship, and freight. GA valuations were thus, to a certain extent, self-correcting. It is possible, of course, that absent merchants based in other ports of call were in a weaker position, but the analysis published by members of the “AveTransRisk” team suggested that valuations were not distorted in this regard.27 If anything, it is likely that all valuations of cargo contained in the database err on the side of generosity, as in many cases this would have produced a lower contribution rate when expressed as a percentage, fraction, or ratio – thus mollifying all participants in the venture.

• 28 URL: http://humanities-research.exeter.ac.uk/avetransrisk/static/data/files/glossary.html.

21Language, terminology, and translation were foreseen issues. Most of the sources were written in different regional Italian languages, whilst the database itself was to be published in English. In line with the guiding principle of data cleanliness, it was decided that terms would be captured exactly as they appear in the sources. This puts the onus for analysis onto the user, who must be versed in the languages in question in order to take advantage of the more language-sensitive data. This is preferable to a situation where input errors are carried deep into the historical record by unsuspecting and perhaps inexpert users. Roles onboard, for example, are kept in the original language rather than “translated”, as these often did not have a direct equivalent. A glossary of these and other terms allows non-speakers an entry point into this data.28

• 29 URL: https://ukdataservice.ac.uk/.
• 30 URL: https://ore.exeter.ac.uk/.

32Once in an accessible format, the data will be deposited into a research repository such as the UK Data Service29 or the University of Exeter’s own ORE repository30. Both systems will make the data available for download and create a unique DOI for referencing. Offering the data in its constituent parts allows easy access to each part of the research and allows other systems to incorporate the data into their own datasets.

• 31 For more information on this voyage, see URL: http://humanities-research.exeter.ac.uk/avetransrisk (...)

34Events in Average sources are usually described with references to nearby islands, towers, and so on. Sometimes there are vague distances in nautical miles (“we were 5 miles north-west of Elba island”), and so on. However, it is hard to determine exactly when a certain event took place, or for how long. For example, how should we locate a storm that started near Sicily, lasted 7 hours and ended near Sardinia? For this reason, the map of events tool is still a work in progress. Currently, only the voyages from 37 Tuscan sources have their events geographically located. An example of event visualisation is shown in Figure 3. The ship Tre Re, with the Flemish “Giovanni di Giovanni” as shipmaster, was freighted to transport wheat from Ancona and Senigallia to Livorno. A few days after setting out they encountered a storm near Cavo Festozzi with north-easterly winds. The crew were set to bailing out and they took in all sail to avoid being pushed onto the land. However, with the storm increasing and the ship threatening to capsize, the shipmaster ordered a jettison of a very large cargo portion after the necessary consultation with the crew. The location of the event A, therefore, was not defined with GPS coordinates, as would be the case today.31

• 32 URL: http://humanities-research.exeter.ac.uk/avetransrisk/voyages/50030/, Voyage ID 50030.

35For similar reasons, we chose not to “draw random lines” to hypothesise the vessels’ routes. For example, although we know that a ship sailed from point A to point B, we cannot know for sure the route it followed. The voyage of the ship called Il Lauro, with a tonnage of 70 lasti (161 tonnes), is a clear example of the unpredictability of maritime routes.32 Shipmaster Simon Sverze of Emden (in modern-day Germany), sailed from Amsterdam to Genoa with a cargo of wheat and rye. Due to a storm, the ship ran aground on a sandbank between England and Ireland. Sverze ordered part of the cargo to be jettisoned to lighten the ship and resume the voyage to Genoa, where he arrived in February 1592. If it were not for the Average event, we would not have known that the ship took a route around Scotland and through the Irish Sea, maybe because of unfavourable winds blowing in the English Channel which were difficult to manage for a vessel that only had square-rigged sails.

• 33 URL: http://humanities-research.exeter.ac.uk/avetransrisk/search/advanced/guide/.

41Various methods of filtering the results are available such as “CONTAINS”, “GREATER THAN” (>), “LESS THAN” (<) and “IS EMPTY”/“IS NOT EMPTY” and brackets can also be used to split the query into smaller parts. There are two types of search fields: text and choice. Text fields allow users to type in the year, word, or phrase to search for. Choice fields allow users to select the enquiry from options presented by a scroll-down list, for example “vessel type” or “cargo type”. Users can build a query around any combination of text and choice fields and add more fields of the desired type. We provided a list of all fields and their field type in the user manual.33

• 34 See A. Iodice & L. Oddo, 2022; A. Iodice & L. Piccinno, 2021.
• 35 According to Douglass North’s definition, transaction costs are the costs of specifying and enforc (...)

42With the advanced search function, the database can be used to investigate, for example, trends in wheat prices, or administrative costs paid by vessels in Genoa for their Average procedures.34 Such costs have recently been studied as a proxy for the overall transaction costs incurred by vessels involved in maritime trade during the so-called “Northern Invasion” at the end of the sixteenth century.35 The following query, for example, also shown in Figure 6, could be used to investigate the evolution of sugar prices and the types of sugar arriving in Genoa between 1590 and 1700, as they appear in Genoese Average sources:

• 36 S. Fenoaltea, 2019.

47First of all, the guiding principles of completeness and data cleanliness increase the demands on inputters. Data cleanliness should, of course, be a sine qua non for all historical projects in order to avoid contaminating the analytic “food chain” at its source. Stefano Fenoaltea, in one of his last essays after a lifetime spent in cliometrics, lamented the tendency of historians to neglect the unglamourous data-collection stage in which vital analytical decisions are taken.36 The result is castles built on clouds, wide-ranging conclusions and impressive-looking econometric analyses based on faulty numbers. However, it should also be recognised that clean data capture usually goes hand-in-hand with more entry fields, as several of the examples related above demonstrate. This does more than simply increase the inputter’s workload. Every additional field increases the complexity of the database and hence the learning curve required for new prospective inputters to become familiar with the entry form. This could reduce the pool of potential collaborators. It also may, ultimately, prove self-defeating, as the increased complexity of the entry form leads inputters to make mistakes, adopt their own idiomatic conventions, or just ignore large sections of the entry form. Lengthy instruction manuals are not an effective safeguard against this kind of human error.

• 37 293 variables in the downloadable version. URL:
• 38 URL: https://riskybusiness.labs.vu.nl/.

48Since data cleanliness is (or should be) non-negotiable, the trade-off is effectively between going deep (a source-oriented database which harvests all potentially useful information from a necessarily limited number of documents) and going wide (a method- or model-oriented database capable of harvesting limited information from many documents). In this respect, database designers would do well to think early about whether their project is going to be an open project to which future researchers may add more data, or a project-specific research tool, whose more restricted design can be guided by research questions (or even whether it is primarily a teaching and dissemination tool). Designers should take account of the resources they have available and pick their strategy accordingly. Slave Voyages is a successful example of a relatively well-funded project that appeals to a large scholarly and public user base and can thus realistically focus on a large number of variables.37 The new Risky Business database, on the other hand, which aims to become a repository for data on historical marine insurance, has wisely decided to focus on just a few variables, notably the insurance policy premium rate, an ubiquitous and analytically useful variable that does not rely on converting currencies.38 The AveTransRisk database could be opened to further collaboration in the future.

• 39 URL: https://anr.portic.fr/en/home/.

51Interoperability aids sustainability because larger datasets are more likely to attract users. Consolidation also allows datasets to be maintained more efficiently. Unfortunately, the AveTransRisk experience demonstrates quite how difficult the creation of such “master databases” might prove to be. The particularities of GA in different early modern ports necessitated the creation of many fields in order to keep data clean, and to deal with various regional and procedural idiosyncrasies. The prospect of simply merging databases – with, for example, the “Portic” project (Ports, and Information and Communication Sciences and Technology: Querying and Visualizing eighteenth-century shipping and trade dynamics in the digital era) – would clearly be a delicate and complex operation rather than a simple case of pouring two datasets into the same digital trough.39

• 40 URL: https://www.geonames.org/.

52More promising in this respect could be a kind of “modular interoperability”, whereby projects draw upon commonly available building blocks that could reduce the time involved in constructing a database and provide a basis for some limited data comparison. The website GeoNames, for example, provides a common fund of coordinates and alternative names that could be drawn upon by future historical databases: rather than creating a port list from scratch, inputters could simply include a reference to GeoNames, allowing for the automatic importing of this data.40 This would also help to reduce human error. Similar online repositories dealing with currencies, premodern weights and measures, or calendars could likewise aid database construction by outsourcing many of difficulties which are currently being confronted “in-house”. A relational structure between online databases may thus be the key to larger datasets in future. It will not, however, provide a substitute for historians making interpretative choices about their data and the way it is presented, or seeking to fully understand the choices that have been made by others in producing large, publicly available datasets.