General Average and All the Rest: The Law and Economics of Early Modern Maritime Risk Mitigation

Abstract

General Average is probably the longest existing maritime trade institution, first used by Greek and Roman sailors in antiquity and still in use today by the global maritime industry. Premium-based marine insurance is much younger. The first evidence for its use is from around 1300 CE. Why did the two institutions originate more than a millennium apart? This timing question cannot be resolved solely by analysing the primary historical records. The answer to this significant question was not written down by contemporaries and will not be found in the texts they left behind. We need theories in order to approach it. While theories about the mitigation of risk might not be the only relevant theories for addressing the question, they are sufficiently instructive to serve as the core of this article.

I would like to thank Giovanni Ceccarelli, Maria Fusaro, Roy Kreitner, Robert Scott, participants in the “Sharing risk: General Average, 6th-21st Centuries” Conference (Genoa, Archivio di Stato, 16–18 May 2019) and anonymous reviewers of the article for valuable comments, references and suggestions and Anna Ashurov and Maya Hay for the excellent research assistance.

General Average is probably the longest existing maritime trade institution, first used by Greek and Roman sailors in antiquity and still in use today by the global maritime industry. Premium-based marine insurance is much younger. The first evidence for its use is from around 1300 CE. Why did the two institutions originate more than a millennium apart? This timing question cannot be resolved solely by analysing the primary historical records. The answer to this significant question was not written down by contemporaries and will not be found in the texts they left behind. We need theories in order to approach it. While theories about the mitigation of risk might not be the only relevant theories for addressing the question, they are sufficiently instructive to serve as the core of this article.

The fastest and least costly way of sending cargos in antiquity was by ships. A geospatial model, created by Walter Scheidel and Elijah Meeks, shows very vividly how travel time and transportation costs rose very steeply in the Roman world once one was getting from the Mediterranean into the surrounding overland routes.¹ We do not have a similar reconstruction of medieval travel time and costs. We can guess that the general picture is similar; the use of natural wind energy, rather than animals that had to be fed, and the larger scale of ships available, compared to packs and wagons, saved time and costs, as they did in the Roman era. The relative differences between sea and overland travel were even higher in the Medieval era due to the absence of well-developed roads and facilities, and the political instability after the fall of Rome.² The use of maritime transportation created an opportunity for substantial trade advantage and hefty profits.

However, maritime trade faced the gravest challenges and the highest risks of all sectors in pre-modern economies. Ships were foundered and wrecked, pirates and foreign rulers could seize goods, cargoes could be damaged, lives could be lost, loans could be negated, agents could abscond, market prices could fluctuate, and demand for goods in the marketplace could dissipate. In a meta-survey of the findings of marine archaeology of the Mediterranean, Andrew Wilson identifies 1646 shipwrecks for the period before 1500 CE.³ This is the lower bound for actual ship loss. More shipwrecks are likely to be discovered as marine archaeology expands its underwater surveys and excavations. Other shipwrecks will never be discovered because they did not survive. Foundering was not the only risk faced by maritime trade; it was the one better captured by modern archaeologists and historians. Jettison, damage to cargo, piracy, business losses were not well recorded but were prevalent.

The environment of maritime trade activities was, in economists’ terms, one of uncertainties, high risks, vast information asymmetries, augmented agency problems, weak enforcement of contracts, and fragile protection of property rights.⁴ Dealing with such a tough economic environment was a foremost institutional challenge for pre-modern contemporary merchants, jurists, and rulers. Maritime trade is where the institutional cutting-edge could be found. This is where new and innovative organizational solutions were designed.

The application of Frank Knight’s theoretical framework, which distinguishes between uncertainty and risk, can take us a long way in answering this question. It focuses on the ability to price risks. As we shall see, Knight’s theory will not take us all the way. I will next show that by introducing Douglass North’s theory of institutional evolution, both the historical story and the theoretical framework get murkier. I will then add Robert Scott’s contracts theory. I will offer a synthesis on both levels, namely on how to overcome the historical puzzle and how to reconcile the theories. After combining the theories into a more comprehensive framework, we can utilize this framework in order to analyse the origins and early evolution of insurance, General Average, and additional maritime trade institutions. I will deal with three of the most renowned of these institutions: the sea loan, the commenda and the business corporation.

This essay does not aspire to offer fully fledged answers to the questions posed above. Instead, it is a demonstration of the value of the theoretical framework offered here in generating research questions and in orientating the historical research. A more systematic and comprehensive application of the theoretical framework offered here for the available historical records is required in order to better base and exhaust the historical insights. This essay originated in a key-note lecture, delivered by an outsider to the field that deals with similar puzzles in different historical contexts, does not report the findings of actual historical research. It is programmatic, setting up the stage in terms of questions. Answers to these questions will have to be provided by historians specializing in the period, such as the scholars taking part in the European Research Council project on General Average (AveTransRisk), led by Maria Fusaro.⁵

Frank Knight on Uncertainty and Risk

Frank Knight (1885–1972) is famed for conceptualizing a distinction between risk and uncertainty. His book in which he created this distinction—Risk, Uncertainty and Profit—is one of the most canonic books in economics and beyond.⁶ Knight explains:

The practical difference between the two categories, risk and uncertainty, is that in the former the distribution of the outcome in a group of instances is known (either through calculation a priori or from statistics of past experience), while in the case of uncertainty this is not true, the reason being in general that it is impossible to form a group of instances because the situation dealt with is in a high degree unique.⁷

Knight identifies three possible states of the world: (1) a priori probabilities which are derived deductively, as in rolling a dice; (2) statistical probabilities which are generated by empirical evaluation of relative frequencies, as in life insurance; and (3) estimates, in which “there is no valid basis of any kind for classifying instances”.⁸ Knight identified risk with points 1 and 2 and uncertainty with 3. From a different perspective, Knight is designated by “risk” situations, in which insurance markets can exist, and by “uncertainty” situations, in which insurance markets cannot.

Knight was not a historian. He does not offer his own historical narrative of the shift of environments from uncertainty to risk. He does not offer a historical account of the origins of marine insurance. One insight is apparent in Knight’s framework. An economic environment cannot move from uncertainty to risk by developing insurance. In an environment of uncertainty, premium-based third party marine insurance (hereafter premium-based marine insurance or simply marine insurance) cannot be developed and offered. Insurance is not the facilitator; according to Knight, it is the outcome. The causation runs the other way around. The birth and further development of marine insurance is a response to a shift of the trade environment from one riddled with uncertainty to one in which uncertainties were already converted to quantifiable risks. Knight does not offer a coherent dynamic theory for the environmental shift. One has to infer the dynamics from his quite static and abstract analysis. Once applying Knight’s theoretical lens, we know where to look for the dynamics. We have to study the informational front. The prohibitive factor that prevents the development of marine insurance is the inability to price maritime risks. The solution was to have more information that would gradually convert uncertainty into risk. Could medieval merchants achieve this? In other words, was increased information exogenous or only somewhat endogenous to the institutional dynamics related to the development of marine insurance? I will get back to these questions later.

Note that, for Knight, insurance is not an invention. It is not the case that, until a certain point in the historical flow, merchants did not understand that risk is a prohibitive or costly hurdle that had to be mitigated in order to expand trade further. It is also not the case that merchants did not know how to organize insurance firms or how to draft insurance contracts. Knight does not consider the need to develop a mathematical understanding of probabilities and their calculations. Even if they fully understood the importance of risk mitigation and had the mathematical, organizational, and legal know-how, they could not invent insurance because they were unable to price the risk premium.

Douglass North on Transaction Costs and Risk Spreading

Let’s juxtapose now Knight’s framework with that of Douglass North (1920–2015).⁹ North was another famed economist. He was economics Nobel Laureate (in 1993) for “renewed research in economic history by applying economic theory and quantitative methods in order to explain economic and institutional change”.¹⁰ North placed on his research agenda institutions, organizations, contracts, and transaction costs, rather than the information needed for pricing. His framework is explicitly dynamic and historical, though admittedly not based on detailed archival research but rather on a stylized narrative. Let’s see whether North’s starting point can provide us with some of the dynamic dimensions and historical specifications that Knight failed to provide. We’ll do it initially with respect to insurance, as this institution was important for both Knight and North.

For North, institutional evolution (in tandem with technological innovation) drives economic development.¹¹ “Institutions exist to reduce the uncertainties involved in human interaction”.¹² The contribution of institutions to economic development is generated by their role in reducing transaction costs and facilitating transactions. Institutional innovations lowered transaction costs, according to North, along three margins: (1) increasing mobility of capital; (2) lowering information costs; and (3) spreading risks. We are interested here in the third.

The institutional evolution that led to the increase in capital mobility included the development of methods for the evasion—and, eventually, the repeal—of usury laws, the development of bills of exchange, better enforcement of contracts, and the development of devices for better monitoring of agents. The lowering of information costs was achieved by the invention of the print, the publication of price and exchange lists, the standardization of weights and measures, a postal system, and more. Insurance and portfolio diversification are the two institutional innovations for spreading and reducing risk, and by this, for reducing transaction costs. The development of marine insurance is a prime example of a risk spreading institutional innovation that lowers transaction costs and enhances economic development.¹³

According to Knight, only after uncertainty is converted into risk does insurance emerge. Do we face here a theoretical contradiction between Knight and North? For Knight, insurance is evidence for the conversion of uncertainty into risk. It is an outcome of the conversion. For North, insurance is a means for transforming uncertainty into risk. Is insurance the institutional innovation that facilitates the conversion of uncertainties into risks? If not, how do we get from a world of uncertainties to a world of risks? So, is this a disagreement between Knight and North as to what came first, the chicken or the egg?

Robert Scott on Contracts Under Uncertainty

While Knight is interested in pricing risk that enables insurance, North is interested in the evolution of the institutions that supply insurance. He takes us in this direction but not all the way. He does not offer a theory for the supply of insurance as an institutional innovation. He does not analyse the organizational and contractual details. Robert Scott, from Columbia Law School, a leading theoretician in the fields of contracts and commercial transactions, can take us the rest of the way from pricing to contracts. Scott has developed over the years, with various co-authors, a theory about the different types of contracts that will be developed and used in different environments.¹⁴ Scott is not a historian. He is interested in the changes that occur as a system moves from the present into the future. This move often involves the introduction of new and experimental technologies, the entrance into new markets and the use of innovative business methods. Scott provides an account of the change in the production of contracts along two continuums, the shift from low uncertainty to higher uncertainty environments, and the shift from thinner (lower scale) to thicker (higher scale) markets. For Scott, uncertainty increases with innovation. Scale increases with maturation. I will use here mainly the first dimension and, along with it, a reverse shift. For me, the general historical shift of markets is from higher uncertainty to lower uncertainty. For Scott, it is the other way around. The difference stems from the fact that I, as a historian, am interested in the longer-term historical shifts and am not focusing only on the initiation stage but also on the maturing stage of any new technology or method.

Scott’s prediction is that, along with the uncertainty to certainty dimension, contracts will evolve from collaborative contracts, that braid formal and informal enforcement and sanctions, to long-term relational contracts that set general standards of behaviour, such as best effort, and finally, to complete bespoke contracts that cover numerous contingencies.¹⁵ The rationale for this evolution is that the lower the uncertainty is, the more feasible it is to draft contractual terms ex-ante, while the higher the uncertainty, the more cost-efficient it is to draft simple contracts and delay the costs to the ex-post litigation stage. Scott’s contribution over Knight is that there are ways to draft contracts even under uncertainty. Contracts can be drafted in a manner that facilitates pricing under uncertainty. He is more concrete, compared to North, about how to do this, namely which types of contracts would fit each environment best. Scott’s theory can guide historical research to look for a specific direction of change over time. However, it cannot provide concrete predictions as to which form of contract is expected to be used at a specific point in time.

By now, we realize that information is pivotal to all three theoretical frameworks. For Knight, information about probabilities is the key. For North, one of the three transaction costs margins, next to risk spreading, is information costs. For Scott, the information determines the suitable contract type. However, Scott also views contracts as a means for generating information. Contracts can require one of the parties to provide information to the other, say by way of requiring agents to report to principals regularly and to provide detailed itinerary and accounts. Scott’s contribution, in this respect, is that his model turns information into an endogenous factor.

Knight, North, Scott and the Origins of Marine Insurance

Let’s use my integration of the uncertainty-risk theoretical frameworks developed by Knight, North, and Scott for explaining the origins and timing of premium-based marine insurance. We’ll start with insurance rather than General Average because it is in the centre of the analysis of Knight and North and yields itself more conveniently for analysis. Premium-based marine insurance first appeared in the Italian city-states of the thirteenth to fifteenth centuries CE Mediterranean.¹⁶ It spread northwards to Antwerp, Amsterdam, and London in the following centuries.¹⁷ Premium-based marine insurance allowed merchants to convert more considerable contingent losses into a smaller fixed charge. How could Mediterranean maritime merchants cross the threshold and move from an environment of unquantifiable uncertainty into an environment of measurable risks?

Knight explains his prime route from uncertainty to risk: “The amount of uncertainty may, however, be reduced in several ways, as we have seen. In the first place, we can increase our knowledge of the future through scientific research and the accumulation and study of the necessary data”.¹⁸ The pricing of the premium depends upon the measurement of probability based on a fairly accurate grouping into classes. Risks did not have to be calculated on the precise individual probability of a single merchant to lose his cargo on the seas, but could also be done at the group level, by finding the proportions of the members of the group which may be expected to lose their cargo. The premium could be set in a heuristic manner, which is not based on a formal concept of probability and mathematical calculations, called by Giovanni Ceccarelli “proto-probabilistic” mathematical thinking or “pre-actuarial” stage in insurance.¹⁹ He shows that the shift from viewing the sinking of a ship as a matter of god’s will or pure luck (that could at most be gambled about) to one that views it as a matter of statistical probabilities (mathematically trained insurance actuaries could price that) was incremental. The insurer had to convince the insured that there was a fairly plausible contention between premium contributions and actual risks, that the insured are bearing a fair share of the burden.²⁰ In a competitive insurance market, the insurer also had to be able to set the price of the risk premium so that it would be sufficiently accurate and competitive, subject the insurance contract to low costs, and enable legal enforcement and effective dispute resolution in the event of disagreements.

A first way identified by Knight for turning uncertainty into risk is by a systematic collection of data about risks. Knight did not develop the distinction between information as non-excludable public good and information as an excludable private good, nor between public gathering and production of information by rulers or municipalities and information gathered by individual merchants. Later scholars did. The development of the Italian city-states supported the information gathering assistance they provided in their hometown, on board ships, and by consuls in overseas ports. It facilitated a more systematic gathering of the fortunes of ships and their cargos and the ensuing losses. This constituted part of the shift from the environment of uncertainty to the environment of risk.

A second way hinted by Knight is by the faster circulation of information due to the increased level of trade. The thickness and scale of trade increased with the commercial revolution. Longer length spans of repeated voyages to the same destinations, higher frequency of voyages, thicker trade networks, and larger commercial and banking family firms gradually led to the accumulation of sufficient information about the a priori probabilities. Knight has in mind the thickness of trade in goods, not of insurance.²¹ As we shall see, Scott has in mind the thickness of the insurance market as an information generator.

A third related way identified by Knight for reducing uncertainty is increased control over the future. Applying this to maritime trade would involve the building of sturdier ships, the elimination of piracy, the setting of international law of naval warfare, the production of better weather forecasts and the use of safer sea routes.²² Once more information about maritime events becomes available through any of the three ways, the environment becomes one of risk rather than uncertainty. Premium insurance soon followed. Transaction costs probably decreased, and economic outcomes improved. The distinction between this third way and the two previous ways of moving from uncertainty to risk is that in the first two, the frequency of negative maritime events is better measured, and in the third, the overall number of negative events is reduced.²³

Information gathering can be endogenous to trade, as increasing frequency of voyages to the same destinations and denser trade networks produce more positive and negative events on which probabilities can be calculated. It can be exogenous to trade, say through more investments by rulers and municipalities in information gathering. Knight reminds us that information is not a free lunch, as the gathering involves costs.

While Knight says that insurance is the outcome of the conversion of uncertainty into risk, North says that Insurance is a solution. North acknowledges that the development of insurance is complicated, yet he does not offer a plausible path for its origins and early development.²⁴ Scott offers such a gradual and relatively plausible and predictable path. The first insurance contracts are predicted by Scott to have been individually tailored. Later insurance contracts are predicted to have been repetitions of previous contracts and eventually boiler-plate standard form insurance policies. The first insurance contracts were likely to cover only one scenario, later contracts a handful of scenarios of damages or loss. More mature insurance contracts were likely to be based on the prediction of numerous likely scenarios and responded to the various eventualities by many if–then clauses. Insurance contracts were likely to evolve into longer and more detailed contracts gradually. It was unlikely that the first insurance contracts included general standards saying that the insurer will cover every damage, every maritime damage or any damage that was not caused by the insured or due to the negligence of the insured.²⁵ All of these predictions can and should be put to historical examination. There is a significant corpus of insurance contracts that survived in notarial and municipal archives in Italian city-states, and they can be examined using these theoretical insights and predictions.²⁶

Four factors complicate the theory-informed study of the origins of insurance based on insurance contracts. First, most contracts did not survive. Second, contracts were also drafted to bypass usury laws and some of their clauses reflect this rather than dealing with information deficiencies. Third, insurers had to find ways to deal with moral hazard problems, which are distinct from pure information problems. Fourth, some of the contractual terms were set in city customs and regulations, and a study of these has to complement the text of the insurance contracts. These complicating factors should not discourage historians of insurance. They could be dealt with in a variety of ways.²⁷

While the development of the form of insurance contracts can be predicted by applying the above theoretical frameworks, the timing of its first appearance is not well postulated by the theory of either Knight, North, or Scott. For Knight, there are two states of the world—uncertainty and risk—and the only way to know that we moved from the former to the latter is by observing the appearance of insurance.²⁸ For North and Scott, the shift is gradual and continuous. One can observe a gradual evolution of insurance contracts. The timing of the appearance of insurance relates to the reduction of uncertainty, which in turn relates to the increased circulation of information. This increase, according to Scott, is partly exogenous—the increased thickness and scale of trade created more information about the probability of negative events. It is also partly endogenous, as the experimentation with insurance contracts and the learning by doing or drafting them generated information. They can, for example, generate information by requiring the policyholder to provide information at the underwriting stage or the claim stage. Such requirements would require the policyholder to collect, record and report information that would not absent the contract. The exact timing in which premium marine insurance was likely to first appear is not postulated by theory. We don’t know exactly how much information of which type is sufficient for pricing insurance accurately enough and for drafting clauses that would cover sufficient if–then contingencies. Even if we knew how much information was theoretically needed, we don’t know to measure how much of the relevant information was actually available for would-be insurers at any point in time. Yet, it will not come as a surprise that the commercial revolution of the eleventh to thirteenth centuries, that took place mainly in the Italian city-states of Venice, Florence, Genoa, Pisa, produced increasing trade information and reduced uncertainty to a level that gave rise to the first insurance contracts.²⁹ We don’t know how far into the revolution, the informational environment became sufficiently thick and certain to give birth to insurance.

The Roman Empire poses a challenge to our Knight-based-analysis, which explains the origins of risk-mitigating institutions, such as insurance, as an outcome of the environmental shift from uncertainty to risk. The enigma addressed in this section is why institutions that were developed in the commercial revolution did not develop accordingly in the previous transition from uncertainty to risk, which occurred during the rise of the Roman Empire. Our discussion, so far, focused on the transition from the Middle Ages to the early modern world. It assumed a linear progression of the level of available information and correspondingly of trade institutions. However, the informational environment of trade in the Mediterranean did not progress linearly from antiquity to the commercial revolution of the Middle Ages, viz. the timing of the birth of insurance. Its progress can presumably be described as an upward sine wave in the Greek and early Roman era, a downward towards the collapse of the Roman Empire that continued into the post-Roman era, and an upwards wave with the rise of Islam and the Italian commercial revolution. The second peak was probably lower than the first (due to the segmentation of the sea to Arab and Latin dominated regions, if not for economic development level as well).

By the first and second centuries CE, the Mediterranean was Mare Nostrum (“Our Sea”), an internal sea of the Roman Empire. Trade networks were thick, the scale of activity was high, Imperial infrastructure was well developed, and looting by foreign rulers and pirates was a non-issue. Information on the loss of ships must have been well known. Probabilities could have been predicted and risks could be priced. The Mediterranean in high antiquity was presumably an environment of risk, not uncertainty, an environment of lower risks than those of the commercial revolution of the late Middle Ages. In Knightian terms, the environment became, with the development of political stability and trade infrastructures, as the Roman Empire prospered, more conducive to insurance. On the face of it, the demand for risk-shifting devices should have been in place. Ships sank in the Roman Mediterranean. This is well known, thanks to marine archaeology, which discovered over the years hundreds of shipwrecks in the bottom of the Mediterranean. Some 1000 such ships were dated to the heyday of the Roman Empire.³⁰

So, the enigma is why didn’t insurance originate in Roman times and had to await the commercial revolution a millennium later? Could it be the case that the Romans did not know how to think about probabilities? Could there have been some obstacles on the way of organizing firms or of drafting contracts? Could there have been Roman substitutes to insurance, which dealt with maritime risks in different methods? The answer to the enigma of why insurance did not develop as early as the Roman Empire is a task for Roman historians. It is beyond the scope of this essay.

Before Marine Insurance—General Average

Let’s use the same theoretical framework for explaining the earlier origins of General Average. General Average is applied in the event of sacrifices of ship cargo or equipment to save the ship and the rest of the cargo. The paradigmatic General Average was initially applied to jettison, which occurred when part of the cargo was intentionally thrown overboard in order to save the ship from sinking, grounding, or capture by pirates, and the ship was indeed saved. Over time, the application of General Average was extended to sacrifice by the ship captain (on behalf of the owner) of sails, ropes, anchor and equipment in order to slow or lighten the ship, and so to save the ship in a storm, to pay pirates a ransom, in coins or jewellery, in order to let the ship and the rest of the cargo go, and other scenarios.³¹ The basic principle was that as the damage was averaged, the loss was shared by all. At least four questions arose: What types of losses, beyond jettison, fall under General Average? By whom should the decision to sacrifice or jettison be made? Which damages should be assessed and shared, and based on which price? Who should take part in sharing the burden of the damages (freight shippers, passengers, crew) and based on what (weight or value)? These complications will be discussed below. General Average deals only with a subset of the potential risks and losses at sea. General Average could not deal with maritime risks as comprehensively as insurance.

General Average is possibly the oldest maritime trade institution still in use today. Its history stretches throughout two millennia, if not more. It is discussed in detail in book 14 of Justinian’s Corpus Juris Civilis, which deals with “The Rhodian Law of Jettison”.³² The Roman jurists, whose texts were collected by the editors of the Digest, go back to the second and third centuries CE. References to Lex Rhodia can be found in first century CE Roman texts, and some scholars believe that a customary Rhodian maritime law existed as early as 600 BCE.³³ Jettison, as a practice, is mentioned in Jonah’s biblical story.³⁴ Thus, with some speculative stance, we can imagine General Average to be 2500 years old. Even a conservative estimate would hold the General Average’s origins to predate that of the premium marine insurance by at least a thousand years.³⁵ Importantly, for our analysis, is that General Average did not develop during the heyday of the Roman Empire when maritime risks were low due to political stability and well-developed economic facilities, but in earlier centuries when risks were high and uncertainties still ubiquitous.

How could General Average function in a world of uncertainty? In case of damage to the ship or cargo, resulting from jettisoning or other voluntary sacrifices of cargo or ship equipment in order to save the rest, all stakeholders proportionally share any losses. Some cargo owners could be more vulnerable than others to jettison, for example, those whose cargo was heavier per value and could have more impact on saving the ship, or those whose cargo was placed, based on loading rules, on the upper deck and could be thrown faster on emergency. General Average better aligned the interests of the various cargo owners so that they would be all supportive of the jettison when effective in saving the ship. General Average also solved ex-ante problems, by making cargo owners more indifferent regarding the location of their cargo on board or whether other cargo on board is more or less valuable per weight than theirs’. The significant advantage of General Average over insurance is that the risk is not allocated to outsiders. The Average was being calculated only between those who had stakes in the ship, cargo owners and passengers. As the risk was not separated and as no one purchased it, there was no need to price the risk ex-ante. All calculations were made ex-post. This solved Knight’s problem.

What about Scott’s contract drafting problem? Is it at all relevant for analysing General Average? Should we understand General Average to be an implicit contract? The question is important for its own sake. It is also relevant in order to ascertain the relevance of Scott’s theoretical framework. We will approach the question using both contemporary and modern understandings of what is contractual. The rules of General Average are known to have been retained in Mediterranean customary laws, in the Digest part of the Justinian Code, and the regulations of Italian city-states. This may create the impression that it was a statute or a regulation. But, in fact, it should possibly be conceptualized as a contractual rule. To be clear, General Average was not a separate, formal, explicit and written contract. One should not expect to find General Average contracts. The question is whether General Average was an implicit background contractual term that was read into freight contracts, even if not spelled up in them.

The conceptualization of General Average as a contractual rule is supported by the location of the rule in the Digest in the sections (Book XIV, Section II) that deal with contracts between shippers and merchants, such as locatio conductio and receptum nautarum.³⁶ General Average could be viewed as either a mandatory rule or as a default rule. When the Romans dominated the entire Mediterranean, the actual effect of the General Average rule was universal; one could presumably not trade on the sea without being subjected to it. Another way of viewing General Average in the Roman Empire is that it does not matter whether one conceptualizes it as a contractual rule or a marketplace regulation; the outcome is similar. In earlier or later periods, it could be viewed as a mandatory rule for every ship departing a port that adopted the rule but not for ships departing other ports. Alternatively, it could be viewed as a take-it-or-leave-it clause in freight contracts, which means that one could not freight without implicitly agreeing to General Average. Yet, it could not be a default clause that some passengers and cargo freighters on a ship would accept, while others would reject by opting out of them. Bear in mind that the fact that General Average was widely followed in the post-Roman Mediterranean ports does not mean that its details were uniform in all these ports. The question that had to be decided was which rules would apply for which ship in each of its voyages. This determination can be understood as being made based on an implicit contractual agreement. Another way of conceptualizing General Average as contract was offered by Levin Goldschmidt. He viewed the germinamento as a contract based on a consultation followed by an agreement to jettison made on the spot on board a ship when the danger becomes imminent. In this version, it is not a law applied to the ship or an agreement made before the departure but rather the decision to jettison that made the General Average enforceable. Andrea Addobbati discusses at length in this volume the history and details of the germinamento and the misunderstandings with respect to it.³⁷ The examination of additional contractual and regulatory aspects of General Average deserves more scholarly attention.

General Average contractual rules were simpler than insurance contracts and specified only one contingency, one set of eventualities, the jettisoning of cargo without the sinking of the ship. In General Average, those whose cargo was not jettisoned had to share the damages. Scott’s framework can be used to explain why General Average contracts were likely to develop before marine insurance contracts. Insurance contracts had to address partial loss as well as total loss, sinking as well as jettisoning, weather-related damages as well as man-made damages by pirates or foreign rulers, and human mistakes made by crew or merchants.

General Average was a good institutional solution for an environment of uncertainty, yet it had its limits. General Average provided no relief in the event of sinking and total loss. In such an extreme event, everybody lost everything; hence, there was no averaging to be made and no stakeholder could share the burden of the loss of other stakeholders.

Having on the same ship cargos that were subjected to General Average and other cargos that were not subjected to it could create unmanageable conflicts of interest at times of trouble. It seems as though ex-ante stakeholders were behind a veil of ignorance, not knowing in advance whose cargo will be jettisoned, so there was no apparent reason for them to reject a freight contract that will include a General Average clause.

Applying the Theoretical Framework for Other Maritime Institutions

Let’s see the value of the theoretical framework developed here for other maritime trade institutions that dealt with risk. We’ll examine first institutional solutions that do not require the crossing of the threshold from uncertainty to risk. We’ll discuss three such institutions in the order in which they appeared historically, the sea loan, the commenda, and the business corporation.

The Sea Loan

Could the sea loan be such solution? In a sea loan, the borrower was exempted from repaying the loan if the ship was wrecked. General Average contracts were less costly to draft and relied on a lower level of knowledge about possible states of affairs. Sea loans were distinct from regular loans in that they provided the ability to separate the allocation of different risks to the two parties. The sea risk, loss of ship or goods at sea, was allocated to the lender, while the business risk, changes in demand and supply and market price fluctuations, was held by the borrower. More specifically, in the case of loss of the ship or goods on the way, in the open sea, either due to drowning or capture by pirates, the borrower was discharged from the debt.

The sea loan originated in antiquity, there is some evidence that it might also have been operative in the commerce of the Phoenician merchant kings of the Levantine coast as early as the second millennium BCE.³⁸ In Athenian law, by the fifth and fourth centuries BCE the sea loan (nautikòn dáneion) was apparently widely used.³⁹ In Roman times, the sea loan (foenus nauticum) was a distinct and well recognized legal category, which was reflected in the Justinian Code.⁴⁰ So, they were a product of early Greco-Roman antiquity and possibly even of the ancient Middle Eastern civilizations. The sea loan was known to the Byzantine Empire, was not accepted by Islamic law and re-emerged in the Latin West with the revival of trade in Italy.⁴¹ The puzzling fact is that sea loans were in use from antiquity, long before insurance. Why did sea loans originate so early?

In theory, the statistical probability of the loss of ships on the high seas had to be known for the sea loan to function. The lender had to know it in order to price the sea risk properly and add the sea risk premium to the interest charged (based on risk-free loan interest and probably also credit default risk premium) for the loan. This probability is unknown in an environment of uncertainty. So, the sea loan seems not to be a well-functioning institution in terms of risk pricing under uncertainty. The Knight-inspired puzzle, as raised in the context of insurance, is reiterated in the context of the sea loan. However, the puzzle why insurance did not appear in the heyday of the Roman Empire when information about maritime losses in the Mediterranean was presumably readily available, does not apply here, because sea loans were available.

However, a few factors may explain why sea loan contracts could be drafted and produced under higher uncertainty than insurance contracts. The drafting of the contract was less costly if we are convinced by Scott’s analysis, in sea loan than in insurance, as it had to cover a narrower set of contingencies. While the maritime risk of full loss at sea was allocated by the sea loan contract to the lender, other contingencies were not covered by the sea loan contract. Regulations, including Roman Catholic usury laws, capped the interest on sea loans in the Latin world from the twelfth century. Thus, the market was not free. Because interest rates could not reflect the risk premium, there was no necessity to know probabilities in order to price them. Yet, it is possible that the response of lenders was through credit rationing, offering loans only for durations and routes in which risk premium could be calculated and the compound was below the cap. Because the risk was only one of several dimensions of the sea loan transaction, the risk premium did not have to be calculated as precisely as insurance contracts in which only the risk was sold.

The Medieval sea loan, the contemporary of the early insurance contracts, was more sophisticated than the sea loan of antiquity. Two variations emerged out of the ordinary sea loan. The first was the bottomry, a loan secured by the ship itself. The second was the respondentia, a loan secured by the goods. Both functioned in an environment of risk, rather than uncertainty, and it indeed functioned differently. The interest rate on sea loans in the twelfth century, which also reflected the sea risk premium, was 25–33% for western Mediterranean round trips, and as high as 40–100% for voyages to the Levant.⁴² The expected profits were high enough to justify this. The sea loan and insurance offered different ways of mitigating risks. The sea loan did not allow the allocation of risk to a third party that was unconnected to the underlining trade transaction and did not yield itself to the spreading of the risk among many, as insurance did. On the other hand, the sea loan did not require precise pricing of the risk, could function with cruder price categories and possibly involved lower transaction costs.

The theoretical framework offered here reshapes research questions concerning the history of the sea loan, introduces new puzzles and offers initial and speculative responses to some of the questions and puzzles. Well-based responses are beyond the scope of the present article.

Commenda

The basic commenda was a bilateral contract involving two parties, the sedentary investing party and the travelling agent. The investing party provided capital in the form of goods and cash and was entitled to a share of the profit. The travelling party provided his labour by travelling with the goods by sea or land to faraway markets in order to exchange the commenda goods with local goods and return with these to the investing party. The itinerary and goods to be bought were set in advance to some extent. The intention was to split the profits upon return based on a pre-agreed basis, typically 25–75%. The commenda was thus an equity-investment contract, specifying investments and payoffs. The commenda was also a labour contract with the travelling party investing labour, expertise, information, contacts and bodily risk.⁴³

What can the theoretical framework of this article inform us about the timing of origins and course of development of the commenda? The first appearance of the commenda was in early Islamic Arabia. There are legal discussions of the commenda (qirad, mudarabah) in Islamic juristic texts by the eighth and ninth centuries. It predated the first use of premium insurance. The commenda first appeared in Italy, in Venice, Pisa and Genoa in the eleventh and twelfth centuries.⁴⁴ Interestingly, the commenda did not exist in the Roman Mediterranean.

Was the commenda well suited to an environment of prevalent uncertainties? In theory, in commenda contracts, the risks are split between the investing party and the travelling party. The investing party risked his financial investment, and the travelling party risked his body and soul, the labour he invested in the project, and his share in the expected profits. The loss at sea was a major risk factor. The investments could vary and so could the share in the profits or losses. The parties could negotiate, in advance, these inputs and outputs based on the expected risk. Risk assessment was influenced by the destination, the length of the venture, the expectations for market prices and profits, and more. Knight’s framework would suggest that an environment of uncertainty cannot give rise to commenda contracts because, in such an environment, risks could not be priced and the splitting of profits and losses could not be negotiated and agreed upon. The shares in the investment and in the payoffs could be viewed as a mode of pricing the risks given the environment. Thus, it could not take place in an environment uncertainty but only of risks.

The Scott layer of the theory may lead to a different postulation. It can explain why the commenda nevertheless originated in an environment of high uncertainties that could not give rise to insurance. The gist of the commenda is that it is a best-effort contract, that the traveller has a sort of fiduciary duty to act to the benefit of the investor, to do business as though it was solely to his benefit. The commenda can be viewed as a formal contract that contains a general standard, whose detailed application was decided ex-post in the event of a dispute, rather than ex-ante specified contingent rules. Unlike in insurance contracts, the specification is done ex-post by the tribunals or courts when settling disputes. In an environment of a less developed legal system, in which third party dispute resolution by professional judges is unavailable, the commenda can be viewed as a collaborative contract that in Scott’s terms braids informal, as well as formal, components. The commenda contract generates information by requiring the travelling agent to deliver itinerary, bills of ladings and accounts of his travels, transactions and profits. Based on these, the investing party verifies whether the traveller shirked or cheated according to prevailing informal norms. The investing party could, in case of a breach, impose a reputational sanction.

Different layers of the theoretical framework used here lead to different, if not contradictory, analysis and conclusions for the development of the commenda. One strand views an environment of risks as a precondition for the commenda. The other sees the commenda as able to function in an environment of uncertainties. The under-determinacy of the theory, with respect to the commenda, may result from the fact that dealing with risks is not its main purpose. The commenda, unlike insurance, was not intended to deal exclusively with risks. It was also, even primarily, an employment or agency contract.

The Business Corporation

The joint-stock business corporation originated only as late as the sixth century. It was, among other things, a response to an increase in uncertainty. This increase did not result from the deterioration in the circulation of information about probabilities in a given environment. The commercial revolution converted uncertainties into risks in the context of the Mediterranean trade. This is the story that explains the emergence of marine insurance. But, this was not the story that gave rise to the business corporation. The relevant development in the sixteenth century was the entrance of Europeans into new trade environments in the Atlantic and around the Cape of Good Hope into the Indian Ocean.

One could revert to the familiar contractual solution, by allocating risks contractually to outsiders. This could be done by way of premium insurance (already regularly used in the Mediterranean), by way of sea loans in which the lender bore the sea risk, or even by way of regular loans in which the lender did not bear the sea risk but bore the risk of insolvency.⁴⁵ But, all of these required attributing probabilities in order to price transferred risks.

The new environment was unfamiliar and adventurous. This was the kind of increase in uncertainty due to innovative business activities along the lines analysed by Scott. Oceanic trade, particularly Asian trade, was thinner and involved higher uncertainties than the well-established Mediterranean trade. Insurance was unavailable for the long Asian voyages that lasted for 2–3 years and often ended in loss.⁴⁶ Because of the lack of prior experience with these long voyages in unfamiliar waters and trade, probabilities of loss could not be estimated, insurance premium could not be priced, and insurance cloud not be offered. The solution was to face the uncertainties by pooling many of them together into a single enterprise. That is, rather than investing in a single ship or a single venture of a few ships sailing together in the same season to the same destination port, investing in an enterprise that operated numerous ships over several years to a variety of ports. The idea of pooling risks together was not altogether new. What was new was the organizational platform used. The innovative organizational solution was the joint-stock business corporation.⁴⁷

The basic theoretical basis for the risk mitigation element of the business corporation can be found in Knight’s theory. He says: “it is simply a matter of an elementary development of business organization to combine a sufficient number of cases to reduce the uncertainty to any desired limits”. And adds: “The possibility of thus reducing uncertainty by transforming it into a measurable risk through grouping constitutes a strong incentive to extend the scale of operations of a business establishment”.⁴⁸ It is interesting to note that, in this sense, Knight is also an institutionalist. For him, the path from uncertainty to risk does not only go through gathering information and calculating probabilities (or increasing control over future probabilities), but also through organizational tools that are not that distinct from North’s.

In our case, the consolidation of numerous ventures, voyages, events and decisions alleviated the need to set the probability for the loss of each ship in each segment of a voyage. The consolidation of several voyages having several ships each made “the law of large numbers” applicable to the first business corporations, the English East India Company [EIC] and Dutch East India Company [VOC]. They handled uncertainties by pooling them together through the longevity offered by the legal personality and the scale offered by joint-stock equity investment. The EIC and VOC did not have to convert these maritime and trade uncertainties into insurable risks in order to enter oceanic trade with Asia.⁴⁹ As the seventeenth century progressed, information was gathered, recorded and processed in the headquarters of these business corporations. The uncertainties were converted into risks, the risks were reduced, insurance became available, and the pooling together of uncertainties was not essential anymore.⁵⁰ By the eighteenth century, the EIC and VOC became rent-seeking monopolies and territorial rulers. Running the trade by smaller merchant houses backed by insurers like the Lloyds of London became potentially more efficient. A political struggle began between the monopolies and new entrants.

Conclusion

Traditional historical analysis is good at reconstructing the pattern of development of maritime trade institutions based on research that looks for the trace of records that were preserved in archives. It is not good at explaining the timing of origins and path of evolution of these institutions. Historians assume, often implicitly, that institutions were invented once societies realized that they are beneficial and learned how to design them. As there are usually no historical records that explain the origins and paths, historians cannot do more than this.

The Knight-North-Scott framework takes us a long way forward in understanding the history of risk mitigation trade institutions. Knight calls attention to the role of information in the shift from uncertainty to risk and the development of insurance. North calls our attention to the role of information in reducing transaction costs and enhancing growth. Scott reminds us that institutions involve contractual drafting and that contracts can deal with information shortage and information generation. By doing this, we realize that the challenge was not to discover how beneficial are insurance and other risk mitigation institutions. The focus, instead, should be on how to solve the informational challenges with respect to risk assessment and pricing and contractual drafting in which without insurance cannot come to life.

The theoretical framework calls for the comparison of different risk-mitigating institutions, in terms of their ability to function in environments of higher and lower uncertainty. Organizing business under uncertainty is different than under risk. The early start and long history of General Average can be explained by the fact that it can function well under uncertainty, even high uncertainty. Once in place, General Average further evolved through contractual refinements of the details of its application. One set of research issues relates to change over time. The details of this evolution can be worked out through a careful historical investigation that is coupled with contract theory.

Information is not only exogenously given but also endogenously generated by institutions. Contracts can be designed so that they will generate the information needed for these contracts to function. The commenda contract is an excellent example of a contract that includes information generation elements. It can function in an environment of uncertainty because it does not rely only on the environment (say the state or the thickness of the trade network) for information but also generates information on the level of the contractual relations.

Another contract theory insight distinguishes between frontend contractual drafting and backend dispute resolution over the contract. The less information is available in advance, the costlier it is to write down a detailed contract that covers the expectations of the principal from the agent in many contingencies and the more cost-efficient it is to write a longer-term relational contract that requires best effort. The commenda is an example of a contractual design that implements this insight in a manner that saves on transaction costs.

Yet, another insight from Scott’s theoretical framework is that the lower the uncertainty, the more feasible it is to draft a complete contingent (if–then) contract. The higher the uncertainty, the more feasible it is to draft contracts that cover a single contingency. This insight fits the fact that sea loan contracts developed long before insurance contracts.

Risk can be mitigated in different ways: allocation, spreading or pooling. We are only beginning to understand under which legal and organizational frameworks and maritime risk environments were risk spread, as opposed to pooled or allocated.

The shift from uncertainty to risk is not well explained by economic theory. The history of organizational solutions for mitigation of maritime uncertainties and risks, from General Average and sea loan to insurance and the business corporation, cannot only benefit from but also contribute to the theory of institutional development more generally.