Medieval Matching Markets
Lars Boerner
Martin-Luther University Halle-Wittenberg,
DAFM King’s College London
Daniel Quint
University of Wisconsin
Draft: April 19, 20191
1The authors thank Chris Turansick for excellent RA work, and Ran Abramitzky, Paul David, OscarGelderblom, Yadira Gonzalez de Lara, Avner Greif, John Hatfield, Randolph Head, Scott Kominers, RamonMarimon, Muriel Niederle, Albrecht Ritschl, Battista Severgnini, and Hermann van der Wee for encouragingdiscussions and comments. We have benefited from comments of participants at conferences at the All-UC Economic History meeting at UC San Diego, the European Economic Association Meeting at BocconiUniversity, the Economic History Society Annual Conference the University of Reading, and seminars at theEuropean University Institute, Humboldt Universitat zu Berlin, Freie Universitat Berlin, London School ofEconomics, Stanford University, and University of Utrecht. Lars Boerner thanks the department of economicsat Stanford University for the hospitality during several visits while this paper was being written.
Abstract
We study the implementation of brokerage regulations as allocation mechanisms in wholesale
markets in pre-modern Central Western Europe. We assemble a data set of 1609 sets of brokerage
rules from 70 cities. We find that brokerage was primarily instituted as a centralized matchmaking
mechanism, with systematic variation in how brokers’ fees were calculated. Brokerage was more
common in towns with stronger economic activities – cities with larger populations, universities,
access to ports and more trade routes, and more politically autonomous cities. Value-based fees were
more commonly used for highly heterogeneous goods, and volume-based fees were more common
for more homogeneous goods. We introduce a simple theoretical model to study the broker’s and
traders’ incentives; we find that this empirical pattern in fees was broadly consistent with the
choices that would maximize total surplus on a product-by-product basis, and that brokerage was
more valuable in unbalanced markets (unequal numbers of buyers and sellers).
Keywords: preindustrial markets, market microstructure, efficient matching
JEL classification numbers: D4, N23
1 Introduction
How markets are organized, and what institutions and mechanisms exist to facilitate trade and
determine the allocation of goods, is an important question in economics. Nobel Prize winner Al
Roth notes, “Traditional economics views markets as simply the confluence of supply and demand.
A new field of economics, known as ‘market design,’ recognizes that well-functioning markets de-
pend on detailed rules” (Roth 2007). But while the tools that modern market designers have at
their disposal – game theory, lab experimentation, computational simulation – may be relatively
new, people have been regulating markets and trying to improve their performance for many hun-
dreds of years. As Roth writes in his new book, “The design of markets, via marketplaces, is an
ancient human activity, older than agriculture” (Roth 2016). Challenges to market design today –
informational asymmetries, strategic behavior, market thinness or congestion problems – are often
problems that were present in earlier times as well.
The design of a market – the specific rules and allocation mechanism implemented – determines
who trades what, how much surplus is generated, and how it is divided among the players in the
market. This eventually has an impact on the development and economic growth of societies.
Thus, understanding early market design can help to relate intentional rational action to long-run
economic development. The importance of the formation of market institutions for the economic
development of societies has been recognized in the literature (North 1981, 1990; Lin and Nugent
1995; Acemoglou, Johnson and Robinson 2005b; Greif 2006; Rodrick 2008). However, the evolution
of market mechanisms has traditionally been seen as an act of spontaneous order in the Hayekian
sense, or a manifestation of the power of the Invisible Hand a la Adam Smith. This is perhaps
surprising; Rodrick (2000) outlined the potential use of mechanism design to better understand the
various forms of market institutions related to economic development in different countries.
The existing literature has considered the formation of market institutions mainly in a bigger
institutional context. Scholars looking into institutional mechanisms have focused on conditions
which enable a smooth working of the price mechanism such as equal access to markets, enforce-
ment of market contracts, and secure property rights in general. Seminal work has been done in
particular by scholars working on the long-run institutional transformation of pre-modern Europe
in a comparative perspective, to better understand the unique European path of economic devel-
1
opment which might have let to the economic divergence between the Western societies and other
parts of the world, and also to investigate more recent institutional transformations using Western
development as a benchmark case. (Important examples are North (1990), North et al. (2009),
Greif (1993, 1994, 2006), and Acemoglou et al. (2005, 2012).)
While these approaches inform us about the formation of institutions guaranteeing equal access
and enforcement of impersonal market exchange, very little has been said about the institutional
detail of exactly how trade was actually achieved: how the aggregation of information worked,
if the search and matching led to the right pairing of buyers to sellers, and how the observed
market organizations affected the welfare gains of market participants and society overall. Historical
investigations indeed show that the allocation of products was a concern (Heckscher 1933, Hibbert
1967); information asymmetry, temporary product shortages, hoarding, and market forestalling
were serious problems at the time. Such historical findings have also been taken up by the market
design literature to motivate current market design problems (Roth and Xing 1994).
This paper aims to fill this gap by analyzing the formation of allocation mechanisms in pre-
modern European towns from a market design perspective. We investigate how societies organized
markets, and whether their market policies were well-designed to have a positive effect on welfare.
We look at one important type of regulated allocation process, the organization of intermediation
in the form of brokerage, primarily in wholesale markets.
We study 231 cities in Central and Western Europe, roughly in the area of the Holy Roman
Empire north of the Alps, during the period from 1200 to 1700. This area and period is particularly
appealing for empirical investigation because local municipalities were typically economically and
politically autonomous. Thus, each city could implement its own types of regulations and allocation
mechanisms, leading to potentially rich variation in detail.
We identify cities with (and without) brokerage regulations, finding 70 cities with brokerage and
1609 sets of regulations. In this sample of regulations, we find certain dominant brokerage designs
with specific combinations of rules. The dominant design was a sort of centralized matchmaking
mechanism: a few licensed brokers specializing in a particular product were given the exclusive right
to offer a service pairing mainly foreign merchants with local buyers, and their behavior was strictly
regulated. Brokers were not allowed to do any business on their own behalf, and were restricted
in what information they could disclose. The brokerage service was open to everybody, rich and
2
poor, foreign and local merchants. Brokers received a pre-defined fee based on the transactions
they generated – most commonly either a fixed fee per unit traded, or a fixed fraction of the sales
price. Empirically, we find that brokerage was correlated with many indicators of economic activity:
brokerage regulations were more prone to appear in larger cities, cities with a university, cities with
sea ports and on more trade routes, and cities with greater political autonomy (and therefore
greater dependence on trade). We also find systematic variation in brokerage design, particularly
in the fee structure, with price-based fees more common for products with greater heterogeneity
and unit-based fees more common for more homogeneous products.
Next, we introduce a simple theoretical model to understand the effects of these rules by focusing
on the incentives facing the broker. We find that the observed patterns in the basis for brokers’
fees was quite rational – price-based fees lead to greater surplus (for both buyers and sellers) when
products have large variation in quality, while unit-based fees lead to greater surplus when products
are more homogeneous and sellers are therefore differentiated primarily by their costs. We also find
that the use of the appropriate type of fees encouraged greater use of the broker by both buyers and
sellers. This gives suggestive evidence that town officials were acting deliberately and rationally in
their design of brokerage rules. Through simulation, we find that the pairings created by brokers
tend to be more efficient than random matching when the market is unbalanced – either buyers
outnumber sellers or vice versa; and that the types of fee structures we observe more empirically
– unit fees for more homogeneous goods, price-based fees for more heterogeneous goods – are the
only ones that perform well in both types of unbalanced markets.
Our insights closely complement the findings of the existing literature. North, Weingast and
Wallis (2009) have identified a process of institutional transition, starting during the pre-modern
times, from a natural state based on personalized power relations to a society with impersonal
open-access order institutions. Similarly, Acemoglou and Robinson (2012) have found the formation
of so-called inclusive institutions. Key in these approaches is the impersonal and equal access to
institutions and the existence of these institutions independently of the current carrier. Our findings
indeed identify such institutions, in the form of official intermediaries who guarantee equal access
not only to the market, but explicitly to an allocation mechanism deliberately designed by the
city-state. The institution is also detached from the person, since a broker was sworn in for some
years and then replaced, or could be dismissed if he did not comply with the rules. In addition,
3
the implemented brokerage mechanism gave the broker an impersonal monetary incentive (based
on the quality and price of the product, not the identities of the traders) to improve the market
allocation.1 The formation and implementation of such a mechanism can be seen as a novelty at the
time, but also as an early point on the timeline of the formation of open access order institutions
discussed by other scholars. Thus it can be seen as an early trial of the fundamental institutional
change suggested in the literature.
In addition, our paper relates to a seminal but controversial finding of Heckscher (1933) on the
origins and early formation of mercantilism. Looking into pre-modern market policy, in particu-
lar market taxation and the organization of staple markets, he concludes that towns implemented
market regulations in the interest of the town society, but also incorporated the interests of foreign
merchants, since towns had to rely on the inflow of consumption goods and raw inputs for produc-
tion. In this way, Heckscher identifies early forms of mercantilism as both protective and liberal,
but strongly with the goal of supporting market transactions and trade flows. However, his claims
of an intentional-acting town or state organizing and channeling market transactions in the interest
of the society had been rather critically received or even disputed by scholars working on related
questions in social science, such as Marc Bloch or Thomas H. Marshall (Magnusson 1994). His
study was based on only a few historical case studies, his analysis on the details of the allocation
process was rather limited, and finally his descriptive theory was seen as rather complicated and
difficult to fully understand. Our findings, based on systematically collected market regulations
and analysis of market mechanisms and policy through the lens of market design, indeed confirm
such intentional welfare-improving policy supporting both local and foreign interests.
Our paper relates to the analysis of market design in a long run perspective. The recent
literature on market design focuses on modern problems (Milgrom 2004, Roth 2008). By extending
this line of research to a long-run historical analysis, we are able to shed light on the identification
and persistent use of specific market clearing techniques in important historical contexts, and begin
to shed some light on a neglected institutional dimension: the long-run evolution of the active design
of market mechanisms. Such a line of research is only now in the process of arising. For instance,
Boerner and Hatfield (2017) study the evolution of financial clearing mechanisms for non-tradeable
1Similarly, Greif (2006) focuses on information flows during the Middle Ages, and elaborates on the transformationof personal reputation-based enforcement mechanisms to an impersonal mechanism in the exchange of merchanttransactions.
4
financial instruments in an environment with limited legal enforcement; Donna and Espin-Sanchez
(2016) study the use of auction mechanisms to allocate water; and van Bochove et al. (2017) study
the use of two-stage auction mechanisms to clear markets for heterogeneous financial securities.
Our results also characterize the environment in which this type of intermediation mechanism
more often appeared. Economic policy-friendly governments (DeLong and Shleifer 1993, Stavenage
2014), university institutions which support education and literacy of the town administraton (Can-
toni and Yuchtman 2014), and trade-geographic advantages (Acemoglou et al. 2005, Bosker et al.
2013) have been linked to prosperous pre-modern European economic development. Based on our
findings, these characteristics can also be related to a higher likelihood of implementing market al-
location regulations. Such an empirical analysis into the connection between deliberately-designed
markets and their environment in such a long run perspective is to the best of our knowledge new
to the literature.
Our paper also contributes to an established literature on intermediation (Gehrig 1993, Yavas
1992 and 1994, Spulber 1996, Neeman and Vulkan 2010, Rust and Hall 2003). This literature argues
that intermediaries improve the welfare of consumers and suppliers by reducing or eliminating
the uncertainty associated with searching for a satisfactory match. Transactions with recognized
centralized intermediaries can supplant decentralized search and bargaining, so that customers and
suppliers avoid the costs of decentralized search. This literature typically studies the decentralized
evolution of different forms of intermediation. Our paper complements these studies by analyzing
a “top-down” implementation of brokers as centralized matchmakers.
Finally, our study contributes to a better understanding of medieval and early modern markets
in general, and of brokerage in particular. Recent studies in economic history have mainly focused
on market access and monopolistic structures in merchant cities (see for example Richardson 2004).
A formal institution analysis and empirical quantification of brokerage is missing so far. The
historical investigation of brokerage institutions has been analyzed rather from a descriptive, holistic
perspective, putting it into the wider historical context (for instance see Murray 2005 or Gelderblom
2012). In a companion paper to this one, Boerner (2016) relates to this historical literature and
studies the multifunctionality of brokerage in detail, where he finds certification and tax collection
as main obligations of brokers in addition to matching; in addition, he relates brokerage to other
5
forms of market mechanisms.2
Our paper is structured as follows. Section 2 discusses the historical economic environment,
and describes the evolution and some common characteristics of the identified brokerage regulations
based on a case study of Frankfurt. Section 3 presents the data set, identifies specific brokerage rules
implemented, and empirically examines patterns in adoption of brokerage regulations, focusing on
the questions of whether or not to implement brokerage and which type of brokerage fees to use.
Section 4 explicitly models broker incentives in a simple theoretical framework, and gives results
about the value of brokerage and conditions favoring one or another type of broker compensation.
Section 5 concludes. Key supporting material is in the attached Appendix; additional empirical
results, theoretical proofs and extensions, and a complete list of sources are in a separate Online
Appendix.
2 City Growth, Trade, and Brokerage
Starting in the late Middle Ages, Western Europe was characterized by economic expansion and
related demographic and institutional changes (Lopez 1976, Bairoch 1988). Driven mainly by inter-
regional trade with merchants traveling to foreign cities, towns started to grow; complementary to
this growth, trade institutions evolved which supported the exchange of products (Verlinden 1965,
Postan 1987, Greif 2006, Dijkman 2011, Gelderblom 2013). In addition to the well-researched
development of institutions enabling enforcement of merchant contracts (Greif 2006), a number
of less systematically studied institutions evolved to support the key elements of trade, such as
information aggregation and the matching between buyers and sellers. In particular, product-
specific spot markets and warehouses for wholesale products, and regulated intermediation in the
form of brokerage, can be frequently documented. Other institutional forms supporting information
flow and allocation processes only developed gradually during this period, and only appeared more
regularly during the the early modern period. (These would include more permanent partnerships in
the form of the first companies, with partners located in geographically separated business offices
(de Roover 1963, pp. 70 ff.; Hunt and Murray 1999, Boerner and Ritschl 2009); weekly public
market price lists (McCusker and Gravensteijn 1991); and centralized market-clearing mechanisms
2A more extensive discussion of the related historical research can be found there.
6
such as sophisticated auction formats (van Bochove, Boerner, and Quint 2017).)
To better understand the form, purpose, and significance of the implementation of brokerage
regulations, we first look into the case of Frankfurt. Frankfurt is an interesting case to study, since
it was one of the most important cities – both economically and politically – during the period
and area of our investigation (Dietz I 1910, Rothmann 1998, Holtfrerich 1999). A historical source
study not only reveals information on the form and function of brokerage, but also on the bigger
historical context and institutional change outlined in the introduction.
Brokerage regulations can be regularly documented in Frankfurt from the middle of the 14th
century until the end of the 17th century (the endpoint of this investigation) and beyond (Schubert
1962). This is typical throughout the area of investigation – brokerage, and brokerage regulations,
persist throughout the 18th century in flourishing towns. The specific regulations and underlying
intermediation mechanisms we can document are very detailed and differentiated. The smooth
functioning of the allocation process must have been a concern of the town officials of Frankfurt,
and they likely established a strong expertise due to their commercial interests. This can be
inferred, for instance, from the organization of brokerage and communication of the regulations to
different foreign merchant groups during fair times (Rothmann 1998, pp. 122ff). It is also reflected
in communication from 1613, when the increasingly-important city of Leipzig asked Frankfurt
officials for advice on how to design their brokerage regulations and mechanisms. Frankfurt sent
their regulations, also referring to other towns like Cologne, Hamburg, and Nuremberg (Moltke
1939, p. 15f.). The brokerage regulations of these towns might have worked as role models for other
cities; in any case, they are fairly representative of what we see in the whole sample (discussed in
the next section).
Brokerage was implemented as a sort of centralized clearinghouse mechanism: a fixed number of
licensed brokers had the exclusive right to match buyers and sellers. A broker was sworn in by the
town and licensed for a specific product or product category. From the middle of the 14th century
we find brokers for wine, meat, horses, and herring (Dietz I, 1910, pp. 379f., Schubert 1962, pp.
28ff.). By the end of the 14th and beginning of the 15th century, brokers for other products can
be observed: among others for cloth and textile industry input goods such as wool, silk, skins and
fur; then spices, metals and iron ware, cattle, property, rents, and bills of exchange.
The regulations restricted the intermediation activities of the brokers to matching buyers and
7
sellers and prohibited brokers from any private business, such as buying and selling for themselves
or partnering with others who did.3 For a successful match they received predetermined fees either
per unit4 or based on the final price.5 The fee structure was specified separately for each product,
and was quite low compared to the selling price.6 This fee could not be changed and was typically
split equally between the buyer and the seller.7 Regulations for some product categories require
brokers to work in groups and share their fees (Schubert 1962, pp. 45f.).8 The use of a broker was
typically not mandatory: only in very few cases over the whole period we study do we find evidence
that a broker had to be used.9 Most of the common rules, as well as the common combinations of
rules, appeared throughout the sample.
Brokers were normally citizens, typically local merchants or producers. During fair times, some
foreign merchant groups were allowed to bring their own (product specific) brokers, who had to
follow the same regulations as the permanent local brokers (Schubert 1962, pp. 37f.).
The brokerage regulations from Frankfurt also reveal a long catalog of penalties for not follow-
ing the rules – in particular, for violating the private business constraint, asking higher fees, or
performing brokerage without being an officially sworn-in broker. In earlier regulations during the
14th century, the punishment was banishment from the city for one year and losing one’s job.10
Later regulations from the 15th century on mainly imposed fines.11 We have no detailed source
material on the application of these punishments, but detailed evidence of the application of similar
punishment mechanisms for brokers can be extensively documented in the city of Cologne during
the same period of time (see Boerner 2016).
The regulations also inform us about the political motivation of town officials in implementing
3See for instance the broker oath from 1465 (Buecher 1915, 213f.)4For example: three Schillinge for one Fuder of wine in 1373 (Buecher 1915, p.325); three Heller for one cow or
swine in 1373 (ibidem, p.221); two Groschen for one ton of herring in 1415 (ibidem, p.241).5For example: four Heller at an outdoor market (six Heller during a fair) from each pound in the price of a horse
in 1360 (Buecher 1915, p. 237); one Groschen for each 100 guilders for bills of exchange around 1450 (ibidem, p.250).6For a general discussion of brokerage fees relative to good prices see Gelderblom 2013.7However, some regulations (particularly early ones) indicated that for some products the seller had to pay the
fee: for instance, the seller paid the entire brokerage fee for horses, and more than half the fee for pigs.8For example, in the wine brokerage fees mentioned above from 1373, wine brokers had to share profits. In
addition, clauses related to information sharing among brokers and “no competition among brokers” can be found inmany towns’ regulations (see Boerner 2016).
9This seems to only regularly have been the case for small retailers (called “Hoecker” (“sitters”)) who boughtfoodstuff products such as eggs, milk, cheese, etc. for daily re-selling on the streets of Frankfurt during the secondhalf of the 14th century – for an example from 1377, see Buecher 1915, p.227. For single observations for a few otherproducts see Schubert 1962, p.57.
10For example in 1357 (Wolf 1969, pp. 11ff.) or 1360 (Buecher 1915, pp.211f.).11For example in 1406 (Buecher 1915, pp.214f.) and 1466 (Wolf 1969, p.356).
8
these orders (Schubert 1962, pp. 69ff.). We find two main types of motivation. During the 14th
century, regulations state that these brokerage regulations have been implemented in the interest
of the town and their citizens. (For an example from 1350, see Buecher 1915, p. 325.) Secondly,
the regulations state that the broker should treat all his customers in an equal way, independent
of whether they are rich or poor, local or foreign. Such statements can be found throughout the
period of investigation; the equal treatment between locals and foreigners, however, we only observe
from the 15th century onward.12 In some regulations, related statements of equal treatment are
explained in more detail when they talk about information sharing between the broker and traders
on either side of the market: the broker was not allowed to inform the buyer if the seller was in
a hurry to sell his goods, nor to tell the seller if the potential buyer was rich or poor.13 He also
could not reveal incorrect estimates of the value of a product (too high or too low) to one side of
the market.
From these sources we can clearly infer concerns about information asymmetries between
traders, which went beyond knowledge of who was selling or buying which types of products,
and concerned product valuations and price preferences of market participants and related con-
cerns about strategic behavior. (If a seller learned from a broker that a buyer could afford to
pay a high price, he would have a stronger incentive to demand a high price; if a buyer learned
about past prices, or that the seller needed to sell his goods quickly, he could understate his own
willingness to pay.14) Thus, the broker was a powerful intermediary, who had the information
and knowledge to influence trade and price formation or even to use the information to do private
business. Regulations and punishment for such behavior indeed show the fear of this abuse, and
the importance of this intermediary function in improving the allocation process.15 Consequently,
the role of brokerage was to support the searching and matching of buyers to sellers, and to solve
12For examples: from 1360 (Buecher 1915, p.211), ca. 1450 (ibidem, p. 249), 1460 (ibidem, p.224).13Such statements can be found for instance in brokerage rules from Frankfurt 1406 and 1465 (Buecher 1915, pp.
211ff. and 213ff.) and 1685 (Beyerbach 1818, pp. 700ff.).14Brokerage regulations from the Alsatian merchant town of Schlettstadt in the early 16th century gave exactly
this second reason for ordering the broker not to reveal information to one side of the market (Geny 1902, p. 988-9).15The conscious implementation, self-reflection and outside perception of these policies can also be seen in various
source material from other important merchant towns such as Cologne or Brugge. For instance in one source fromCologne, an expert probably ordered by the city of Cologne evaluates different market making activities and regu-lations, including brokerage regulations. The report concludes that the brokerage regulations in use are good andshould be kept (Stein II, 1893-5, pp. 565f.). In a letter from the Hanseatic League to Brugge in 1438, merchantswho had earlier left Brugge were negotiating over returning to Brugge to do business. Among other demands, theywrote that they only would come back if the city could guarantee that brokers did no private business for themselves(Hoehlbaum et al. 1876-1939, Hansisches Urkundenbuch, VII n. 389 § 5).
9
the information asymmetry problems between both sides of the market in a “fair way.”
3 Brokerage regulations
Having discussed brokerage in Frankfurt, we proceed with a more comprehensive study. Our study
covers towns in the area of Central Western Europe, basically the outline of the Holy Roman
Empire north of the Alps at its largest, as well as eastern neighboring cities in the kingdoms of
Poland and Hungary. Following Bairoch et al. (1988), our investigation covers the years 1200 to
1700, and includes all cities in this area which had at least 5000 inhabitants at some point during
this period. This means 231 towns were considered, in 70 of which brokerage could be identified.
Figure 1 shows all towns considered, and distinguishes those in which brokerage regulations were
found from the others.
The data are compiled from edited and non-edited sources based on several thousand pages of
source material, which have been translated and analyzed by us from different mainly medieval
Germanic dialects. We analyzed all edited sources available for the area and period of investiga-
tion, as well as complementary archival material mentioned in the edited documents or secondary
literature. Additionally, we checked for documented archival material in all cities in the area of
investigation mentioned in Bairoch et al. (1988). The composition of our sample of brokerage
regulations thus reflects the random survival and accessibility of the sources, not a conscious biased
selection on other bases. (A complete list of all sources can be found in the Online Appendix.)
3.1 Existence of Brokerage Regulations
We begin by investigating which cities instituted brokerage at all, and how they compare with the
cities that did not. This gives us some initial quantitative insights beyond historical narratives
about which towns implemented such market regulations, and perhaps why. In particular, we
analyze if the existence of brokerage regulations can be linked to other economic activities.
To create a variable which approximately indicates whether brokerage was in use in a town at
each point in time, we divide our period of investigation (1200-1700) into 50-year intervals, creating
ten time windows for each of the 231 cities and therefore 2310 potential “observations”. For a time
period in which a particular city had not yet been founded, or had not yet achieved a population
10
size to be captured in the Bairoch city population statistics, we drop the observation, leaving us
with 1823 city-time-period combinations in which to examine whether brokerage was used.16 We
create a binary variable for each of the 1823 city/time observations, which takes the value 1 if we
find brokerage regulations mentioned at least once in that time period, and 0 otherwise. Of the
1823 city/half-century pairs, 225 contained an observation of brokerage rules, thus about 12 percent
of all observations.17
We then examine whether this variable is correlated with various demographic, political, and
trade-geographic variables which have been used in the past by scholars to determine economic
activity and development in the period and area of investigation. (Explanatory variables are defined
in detail in the Appendix.) To determine sign and significance of these statistical relationships, we
run regressions (both OLS and probit) using the existence of brokerage as the dependent variable
and this set of explanatory variables.
Table 1 shows the descriptive statistics of the variables. For each binary variable, it also includes
the fraction of observations where brokerage was found when this variable is equal to 1, and for the
other variables, the average of the variable in the observations where brokerage was found. Thus,
Table 1 shows that brokerage was found in 29% of observations of cities containing a univeristy,
much higher than the overall sample (12%); and that the average population in observations where
brokerage was found (17,040) was almost double the average population in the overall sample
(8,820). Table 2 depicts the output of the related regressions, with a binary variable for the existence
of brokerage regulations as the dependent variable. Columns (1)-(6) use probit regressions, and
incorporate various sets of variables and controls, starting with institutional variables, next adding
trade-geographic variables such as port access and the number of trade routes leading to a city,
then population size, and finally an extra check on past institutions (whether a town was originally
a Roman town); column (7) shows results for OLS regression with the same variables as column
(6), for comparison. (OLS results corresponding to all six probit specifications can be found in
the online appendix.) Error terms are clustered by city, and also by century for columns (4)-(7).18
16In an alternative specification, we replace the observations without population information with an arbitrarysmall population size of 500 inhabitants; the results do not change (available on request).
17The source investigation resulted in a total of 1609 observed sets of brokerage rules distributed among 225city/half-century pairs. A detailed analysis of these regulations follows in the next section.
18Clustering of error terms of observations within the same century is necessary when controlling for city populationsize, because we report brokerage observations in fifty-years-intervals, while Barioch reports population only everyhundred years, requiring interpolation of city size. In addition to probit and standard OLS regressions, we also ran
11
Table 1: Summary Statistics – Existence of Brokerage
Variable Mean Std Dev Min Max % withBrokerage
Mean withBrokerage
Dependent VariableExistence of Brokerage 0.12 0.32 0 1
Explanatory VariablesFree Imperial City 0.19 0.39 0 1 25%Bishop 0.22 0.41 0 1 18%University 0.07 0.25 0 1 29%Hanseatic League 0.18 0.39 0 1 12%Roman City 0.13 0.34 0 1 22%Water (any port) 0.51 0.50 0 1 15%Sea Port 0.14 0.34 0 1 26%Number Major Trade Routes 3.2 2.2 1 9 4.66Population (thousands) 8.82 9.93 1 127 17.04
Control VariablesYear 1452 139 1200 1700 1464Longitude 9.66 5.04 2.88 52.52 8.69Latitude 50.53 2.11 36.64 54.78 50.42
Notes: The total number of observations is 1823. Institutional variables change over time, geographic and trade-geographic variables do not.
Note that many of these explanatory variables are themselves endogenous, and we interpret the
regression results as being more descriptive than causal.
We find clear positive relationships between the existence of brokerage regulations and a number
of key variables established in the existing literature as being indicative of the level of economic
activity. First, we find a positive, statistically significant relationship between the existence of
brokerage and the population size of a city. Population size has been identified as a good proxy
variable for economic growth and development in pre-modern European towns (DeLong and Shleifer
1993, Acemoglou et al 2005). Brokerage also correlates with the trade-geographic location of a town.
Cities with access to the sea, and with more trade routes reaching the town, were more likely to
implement brokerage regulations. The importance of trade geography for economic development
has been for instance identified by Bosker et al. (2013), and the importance of proximity to the
sea coast, in particularly the Atlantic coast, by Bairoch (1988) and Acemoglou et al. (2005).
We find a significant relationship between the implementation of brokerage regulations and the
simple significance tests for the correlation between brokerage and each dependent variable, and found very similarresults.
12
Table 2: Regression Results: Existence of Brokerage Regulations, 1200-1700
(1) (2) (3) (4) (5) (6) (7)
Dependent Variable: Existence of BrokerageRegression Type: Probit Probit Probit Probit Probit Probit OLS
Free Imperial City 0.17*** 0.16*** 0.09*** 0.07*** 0.07*** 0.08*** 0.11***[0.04] [0.05] [0.03] [0.02] [0.02] [0.02] [0.04]
Bishop 0.03 0.04 0.00 -0.03 -0.02 -0.04 -0.02[0.03] [0.03] [0.02] [0.03] [0.03] [0.03] [0.04]
University 0.20*** 0.20*** 0.12** 0.07*** 0.08*** 0.07*** 0.10***[0.06] [0.07] [0.05] [0.05] [0.02] [0.02] [0.04]
Hanseatic -0.01 -0.01 -0.03 -0.04 -0.04 -0.04 -0.03[0.03] [0.03] [0.02] [0.03] [0.03] [0.03] [0.03]
Roman city 0.04 0.05[0.03] [0.04]
Water (any port) -0.02 -0.01 -0.01 -0.01 -0.02[0.03] [0.02] [0.02] [0.02] [0.03]
Sea port 0.28*** 0.15*** 0.16*** 0.16*** 0.21***[0.08] [0.03] [0.03] [0.03] [0.04]
Number trade routes 0.03*** 0.02*** 0.02*** 0.02*** 0.02***[0.01] [0.01] [0.01] [0.01] [0.01]
Log(Population) 0.08*** 0.08*** 0.09***[0.02] [0.02] [0.02]
Population Quintile 2 0.09**[0.04]
Population Quintile 3 0.13**[0.05]
Population Quintile 4 0.15***[0.05]
Population Quintile 5 0.21***[0.05]
Year 0.00 0.00 -0.00 -0.00 -0.00 -0.00[0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Longitude -0.01* -0.01*** -0.00** -0.01*** -0.00*** -0.00*[0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
Latitude 0.00 -0.01 -0.01* -0.01* -0.01* 0.01[0.01] [0.01] [0.01] [0.01] [0.01] [0.01]
Observations 1823 1823 1823 1823 1823 1823 1823No. of City Clusters 224 224 224 224 224 224 224No. of Century Clusters . . . 5 5 5 5Pseudo R-squared 0.08 0.10 0.20 . . . 0.25Log (pseudo-)likelihood -615.2 -609.4 -529.6 . . .Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors in brackets, clustered by city code forcolumns (1)-(3), by city and century in columns (4)-(7). Results report marginal effects for probit regression;column (7) gives OLS coefficients for the same set of explanatory variables as column (6) for comparison.Column (1) measures institutional effects; column (2) adds geographical and time control variables; column(3) adds trade-geographic effects; columns (4) and (5) add population effects; column (6) adds a control forthe past institutional effects of Roman towns.
13
political organization of a town. Autonomous cities such as Free and Imperial cities more commonly
implemented brokerage regulations than so-called territorial cities, which were governed by a duke
or bishop. This is in line with previous scholars (DeLong and Shleifer 1993, Acemoglou et al. 2005b,
Bosker et al. 2013, Stavenage 2014), who have argued that autonomous cities were particularly
active economically due to their need to support local production and trade, but also due to the
incentives of the local participative governments, which were at least partly represented by local
craftsmen and merchants. In contrast, cities ruled top-down by either dukes or bishops could rely
on other income streams (rents and taxes), and had less incentive to promote local trade activities.
(Cities classified as neither “Free Imperial” nor “bishop” were ruled by a territorial duke who lived
outside the town; the dummy variable for territorial cities is excluded from the regressions, so
coefficients on the other city types can be thought of as relative to the baseline of a territorial city.)
Finally we find that the existence of a university in a town positively relates to the implemen-
tation of brokerage regulations. This is in line with Cantoni and Yuchtman (2014), who argue that
universities produced administrative experts who were able to read and write and in this way could
be instrumental for the town in implementing market regulations.19
We do not find that belonging to the Hanseatic League had a differential effect on a city
instituting brokerage. This might seem surprising at first, since we would expect that cities who
organize their trade political interests abroad would also organize the markets inside the town walls.
However, scholars of the Hansa (Dollinger 1966, Friedland 1991) argue that cities belonging to the
Hanseatic League were rather heterogeneous and not all of them were so actively involved; thus,
membership was not such a strong indicator of trade activities per se.
(The results also suggest that we are not simply finding the results of a selected sample of
surviving source material based on the population size of a city. We find a large number of significant
effects beyond city size, which are related to other institutional and geographical characteristics.
For example, we more frequently found brokerage regulations in Free and Imperial cities than
in bishop-led cities, despite the fact that they were smaller on average and that towns with an
ecclesiastical institution and leader typically archived materials over a longer period of time.)
19A direct link between brokerage rules and historical roots of legal codification based on a Roman inheritagecannot be supported. Controlling for a town’s Roman origin does not affect the results, and the variable itself is notstatistically significant.
14
3.2 Specific Regulations Implemented
Next, we return to the sample of 1609 sets of brokerage regulations we found in 70 cities (out of
231 cities investigated) and look at the details of how brokerage was organized. The sample allows
us to identify the main brokerage rules implemented, as well as frequently-used combinations of
rules. This enables us to better detect how towns framed and constrained the roles and incentives
of merchants and intermediaries.
Table 3 depicts the statistics of the individual and combinations of regulations. It shows how
frequently different types of rules appear in total, and in how many distinct towns.
Table 3: Summary Statistics – Brokerage Rules
Rule or Rules Observations Distinct Towns
Individual Rules
Only licensed brokers (a) 820 48Private business constraint (b) 480 34Fixed fees (c) 695 50
Unit fees (d) 437 41Value fees (e) 235 35
Forced brokerage (f) 81 16
Combinations of Rules
Matchmaking (a+b+c) 353 28Matchmaking with unit fees (a+b+d) 231 22Matchmaking with value fees (a+b+e) 116 20
Matchmaking without private business constraints (a+c) 125 22Matchmaking without fixed fees (a+b) 30 15Forced matchmaking (a+b+c+f) 44 6
The most commonly-used rule gave licensed brokers the exclusive right to match buyers and
sellers (only licensed brokers); this was found in 820 observations from 48 towns.20 Brokers were
often prohibited from conducting private business (private business constraint) – this prohibition
20Other forms of intermediation were generally not allowed. Some early sources documented brokers sharing thisprivilege with innkeepers – in particular, this was seen in the area of the Netherlands and Belgium (see also Boerner2016), for example in Brugge (Gilliodts van Severen 1881, Gelderblom 2013). Furthermore, there is evidence thatduring the 17th century in fast-growing cities such as Amsterdam (Noordkerk 1748, vol.2, pp.1060-3), Hamburg(Beukemann 1912, pp.545-61), Leipzig (Moltke 1939, p. 14f.), and Nuremberg (Roth 1802,p.338), private intermedi-aries who acted as matchmakers were temporarily tolerated and then forbidden again.
15
was found in 480 observations in 34 towns. This constrained the brokers from making profits, acting
as a private buyer, being in open and silent partnership with others, or working on commission
for non-present merchants. In some cases brokers were also forbidden from being a host for their
customers.
Brokers typically earned a pre-defined brokerage fee (fixed fees), which came out of the price
paid in the transaction, and was only paid after the proposed sale was agreed to by the merchants
and the transaction was completed. This was found in 695 observations in 50 towns. The most
common was a fixed fee per unit of goods traded (unit fees). Also common were fees which depended
on the price paid. These were most commonly a simple percentage of the transaction price, but
were sometimes nonlinear or step functions; we code all of these as value fees.21
Use of a broker was typically optional; the obligation to use a broker (forced brokerage) was
only found in 81 observations in 16 towns. Such an obligation was typically only temporary, and
for selected goods.22
We can also look at the combinations of rules found together in each observation, and observe
a similar pattern as discussed for Frankfurt. In 353 sets of regulations from 28 towns, the town
limited brokerage to a few licensed brokers, imposed the private business constraint, and set a
predetermined fixed fee (either unit or value-based). Since the broker could not trade on his own
but could only facilitate matching of buyers to sellers, we can think of their role as that of a
centralized clearinghouse and matchmaking mechanism. We refer to this combination of rules as
matchmaking ; and more specifically, as matchmaking with unit fees or matchmaking with value
fees.23 The prevalence of this particular matchmaking mechanism can be further documented
when looking at one of the most important trade routes of the area and period of investigation, the
Rhine-Main-Meuse-Scheldt area (Irsliger 2010). Figure 2, on the left hand side, illustrates the use
of the matchmaking mechanism in towns along this trade route; on the right side, it shows the use
of this mechanism for one specific product category, wine, from 1350-1400.
21The number of observations of fixed fees do not exactly equal the sum of unit and value fees, since a few sourcesreveal information on the existence of fixed fees but do not inform us of their exact nature, and in some cases we findboth types of fees in one observation.
22A permanent obligation to use a broker as an intermediary can be only documented in Brugge (see Gilliodts vanSeveren 1881).
23Again, the numbers in Table 3 for the matchmaking mechanisms with different fees do no exactly aggregate upto the category matchmaking, since in some observations we are only informed about the fixed fees, but not aboutthe specific form.
16
In a smaller number of observations (125 regulations in 22 towns), brokers were not prohibited
from conducting private business on their own behalf, but the other regulations were the same –
brokerage was limited to a small number of licensed brokers, merchants could choose whether to
use a broker, and brokers received fixed unit or value fees when the trade had taken place. We
refer to this combination of rules as matchmaking without private business constraint. While this
combination of rules has a flavor of intermediaries who act as market makers (i.e., brokers could
have acted as re-sellers on a permanent basis), no such activity of official brokers can explicitly
be documented from the sources.24 In another, less common combination of rules, merchants were
required to use a broker, while the other regulations (only licensed brokers, no private business,
and fixed unit or value fees) remained the same; we refer to these as forced matchmaking. Even
less common were regulations which did not specify a fixed level of fees (matchmaking without fixed
fees), but where the other dominant rules remained the same.
3.3 Product Categories and Brokerage Fee Basis
One important variation in the most frequently documented matchmaking design was the choice
between unit and value fees. Since private business was prohibited, brokerage fees were the direct
income streams for the brokers, and consequently determined the incentives they faced; as we will
further investigate in the theory section, this can have a large effect on the broker’s behavior and
therefore on the outcomes achieved.
As exemplified in the case of Frankfurt, brokerage regulations were for particular products or
product categories. The products covered were basic foodstuffs such as fish, grain, wine and beer,
cattle and meat, and oil and fat; finished cloths or input goods for the textile “industry” such
as raw textiles (wool, linen, fustian, etc.) or fur, skins, and leather; spices and similar products
(in particular coloring products), construction material, metals, financial products (including gold
and silver), and property (land and houses). All of these categories could be found through the
whole period of investigation, typically in towns along trade routes for these type of products, as
for example shown in Figure 2 for wine brokerage in towns along the Rhine-Main trade route.
Table 4 gives some descriptive statistics for each product category. The first two columns show
24Outside our period of investigation, during the 18th century, official brokers who acted as market makers can bedocumented – see van Bochove (2013) and Santarosa (2013). Thus, the relaxation of the private business constraintwhich we document, particularly later in our sample, might have been the starting point for such a movement.
17
Table 4: Summary Statistics – Product Categories and Unit vs Value Fees
Product Obser- Towns Fixed % Unit Matchmaking w % Unitvations Fees Fees Fixed Fees Fees
Finance 103 24 42 7% 18 11%Property 47 16 24 8% 12 0%Horses 83 23 30 7% 19 0%Wine and beer 124 33 56 95% 32 97%Grain 78 26 38 79% 22 86%Fish 89 22 33 94% 15 93%Cattle and meat 50 18 19 89% 8 88%Oil and fat 61 23 50 98% 21 95%Construction material 78 21 50 82% 22 100%Metal 61 20 37 76% 19 74%Spices 77 27 52 67% 24 67%Raw textile 104 28 56 43% 31 39%Fur, skin and leather 81 24 43 65% 23 57%Cloth 108 32 45 36% 24 13%
Total 1144 64 588 63% 304 60%
how many times, and in how many distinct towns, brokerage rules were observed for each product
category. The last four columns relate to the choice of unit versus value-based brokerage fees.
Column three shows how many sets of regulations for a product category specified either unit or
value fees (fixed fees); column four shows the fraction of these observations which specified unit fees.
Column five shows how many observations for a product category specified fixed fees in combination
with the other regulations in the dominant combination we labelled matchmaking (brokers having
the exclusive right to act as intermediaries, being barred from private business, and brokerage not
being mandatory); column six shows the fraction of these observations which specified unit fees.
So for example, brokerage rules for financial products were observed 103 times, in 24 towns. In
42 of these observations, either unit or value fees were specified, but in only 6 of them (7% of 42)
were they unit fees. In 18 observations, the matchmaking combination was used with fixed fees; in
just 2 of these observations (11% of 18) were the fees unit fees. Thus, while unit fees were slightly
more common than value fees in the overall sample, financial products were much more commonly
traded with value fees.
Table 5 gives regression results to establish whether product-by-product patterns in the type
18
of fee used are statistically significant, and persist when we control for city size, time, geography,
political institutions, and trade geography as in the earlier regressions. Regressions use the 588
observations of fixed fees, with the dependent variable being a dummy for whether unit fees (as
opposed to value fees) were used. Columns (1)-(4) give OLS results, with different combinations of
controls; columns (5)-(8) report marginal results for the corresponding probit regressions. A nega-
tive coefficient therefore indicates that unit fees were more common for a given product, a negative
coefficient that value fees were more common. Error terms are again clustered by city, and (for
the OLS) by century when controlling for population. Table 6 shows analogous results for the sub-
sample of observations where fixed fees were found in combination with the other “matchmaking”
combination of rules.
Looking at the various product categories depicted in Tables 4, 5, and 6, three products were
traded much more frequently via brokerage with value fees: horses, property, and financial products.
(These have small percentages in columns 4 and 6 of Table 4, and statistically significant negative
coefficients in Tables 5 and 6.) All these products can be thought of as having heterogeneous
quality characteristics and likely heterogeneous tastes among buyers. Horses and properties are
very idiosyncratic products by nature, and especially for the latter there would be strong differences
in size. Financial instruments, for example bills of exchange, were also very heterogeneous, as
preferences for them depended very much on the participants and the place the bill was drawn
on. The limited tradeability of bills of exchange made the valuation even more heterogeneous (van
der Wee 1963, North 1981, Munro 1994). Of course we can also expect big differences in size for
financial products.
At the opposite extreme, many product categories were for basic consumption goods like grain,
wine and beer, fish, cattle and meat, oil and fat, which were rather homogeneous. (Only a small
fraction of wine consumed was high quality wine, where we might expect more variations in value
– see Rose 2011 and Matheus 2004.) Most of these products were traded more often with unit
fees, and the statistical relationships between product and fee structure are again mostly positively
significant. Similarly, the rather homogeneous raw input and construction materials such as wood,
bricks, and metals were mainly traded via unit fees.
Finally, there were some products which were commonly traded via both unit and value fees.
Accordingly, we cannot assign a statistically significant relationship between these products and
19
Table 5: Regression Results – Determinants of Fee Structure, 1200-1700
(1) (2) (3) (4) (5) (6) (7) (8)
Dependent Variable: 1 if unit fees, 0 if value fees (sample is brokerage rules with fixed fees)Regression: OLS OLS OLS OLS Probit Probit Probit Probit
Finance -0.47*** -0.47*** -0.48*** -0.47*** -0.50*** -0.54*** -0.55*** -0.55***[0.08] [0.08] [0.07] [0.09] [0.13] [0.13] [0.13] [0.13]
Property -0.42*** -0.42*** -0.42*** -0.42*** -0.44*** -0.51*** -0.50*** -0.50***[0.09] [0.09] [0.08] [0.13] [0.18] [0.16] [0.16] [0.16]
Horses -0.49*** -0.52*** -0.52*** -0.53*** -0.45** -0.52*** -0.54*** -0.54***[0.14] [0.18] [0.14] [0.14] [0.20] [0.18] [0.17] [0.16]
Wine and Beer 0.29*** 0.29*** 0.28*** 0.28*** 0.34*** 0.34*** 0.33*** 0.33***[0.06] [0.06] [0.06] [0.04] [0.04] [0.04] [0.03] [0.08]
Grain 0.21** 0.20* 0.19* 0.20 0.26*** 0.23*** 0.22*** 0.22***[0.10] [0.10] [0.10] [0.14] [0.06] [0.07] [0.07] [0.07]
Fish 0.34*** 0.34*** 0.34*** 0.32*** 0.34*** 0.33*** 0.33*** 0.33***[0.08] [0.08] [0.08] [0.14] [0.04] [0.03] [0.03] [0.03]
Cattle and meat 0.22** 0.22** 0.21** 0.20** 0.26*** 0.24*** 0.23*** 0.23***[0.09] [0.09] [0.07] [0.08] [0.07] [0.07] [0.07] [0.07]
Oil and fat 0.34*** 0.33*** 0.32*** 0.32*** 0.37*** 0.36*** 0.36*** 0.36**[0.06] [0.06] [0.06] [0.010] [0.04] [0.03] [0.03] [0.03]
Construction material 0.23* 0.22* 0.21* 0.21 0.27** 0.24** 0.23** 0.24***[0.12] [0.12] [0.12] [0.13] [0.08] [0.09] [0.09] [0.09]
Metal 0.20*** 0.08 0.06 0.06 0.20*** 0.17* 0.16* 0.16*[0.08] [0.11] [0.11] [0.13] [0.08] [0.09] [0.09] [0.09]
Spices 0.02 0.00 -0.01 -0.00 0.14* 0.09 0.08 0.11[0.08] [0.08] [0.08] [0.07] [0.07] [0.09] [0.09] [0.09]
Raw textile -0.17* -0.18** -0.18** -0.26** -0.04 -0.10 -0.10 -0.10[0.08] [0.08] [0.08] [0.11] [0.15] [0.16] [0.16] [0.16]
Fur, skin and leather 0.02 0.03 0.02 0.02 0.11 0.09 0.09 0.09[0.10] [0.10] [0.11] [0.11] [0.12] [0.12] [0.12] [0.12]
Cloth -0.31** -0.31** -0.32*** -0.32** -0.19 -0.25 -0.27 -0.27[0.19] [0.12] [0.12] [0.12] [0.19] [0.20] [0.19] [0.19]
Geography and time YES YES YES YES YES YES YES YESInstitutions YES YES YES YES YES YESTrade geography YES YES YES YESPopulation YES YES
Observations 588 588 588 588 588 588 588 588City Clusters 44 44 44 44 44 44 44 44Century Clusters . . . 5 . . . .(Pseudo) R-squared 0.38 0.40 0.40 0.41 0.55 0.57 0.60 0.60Log (pseudo-)likelihood . . . . -250.5 -243.2 -240.9 -240.9Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors clustered by city code (and by century for column (4))in brackets. Columns (1)-(4) give OLS results for unit (as opposed to value) fees for all products. Column (1) controls fortime and geographical coordinates, (2) incorporates the institutional control variables, column (3) the trade geographicvariables, and column (4) population. Column (5)-(8) report marginal effects for corresponding probit regressions.Results for the regression in column (8), but with standard errors clustered by century, can be found in the onlineappendix, in column (6) of Table 11.
20
Table 6: Regression Results – Determinants of Fee Structure, 1200-1700
(1) (2) (3) (4) (5) (6) (7) (8)
Dependent Variable: 1 if unit fees, 0 if value fees – smaller sample of “matchmaking” observationsRegression: OLS OLS OLS OLS Probit Probit Probit Probit
Finance -0.46*** -0.47*** -0.46*** -0.46*** -0.53** -0.71*** -0.72*** -0.72***[0.17] [0.14] [0.15] [0.15] [0.19] [0.14] [0.14] [0.14]
Property -0.56*** -0.59*** -0.56*** -0.56*** . . . .[0.17] [0.17] [0.17] [0.17]
Horses -0.67*** -0.70*** -0.68*** -0.67*** . . . .[0.14] [0.13] [0.15] [0.15]
Wine and Beer 0.26** 0.26** 0.28** 0.28** 0.33*** 0.32*** 0.32*** 0.33***[0.11] [0.11] [0.11] [0.11] [0.03] [0.05] [0.05] [0.05]
Grain 0.20 0.23 0.24 0.24 0.29** 0.24*** 0.22** 0.22***[0.14] [0.16] [0.16] [0.16] [0.10] [0.09] [0.09] [0.07]
Fish 0.34* 0.33* 0.35* 0.35** 0.32*** 0.29*** 0.26*** 0.26***[0.17] [0.14] [0.17] [0.17] [0.07] [0.05] [0.06] [0.06]
Cattle and meat 0.15 0.15 0.14 0.14 0.18 0.01 0.01 0.02[0.14] [0.08] [0.14] [0.15] [0.17] [0.22] [0.23] [0.23]
Oil and fat 0.35** 0.33** 0.35** 0.35** 0.34*** 0.31*** 0.28*** 0.28***[0.16] [0.16] [0.16] [0.16] [0.07] [0.05] [0.06] [0.06]
Construction material 0.42** 0.42** 0.42** 0.42** . . . .[0.20] [0.20] [0.20] [0.20]
Metal 0.12 0.11 0.11 0.11 0.19 0.15 0.16* 0.16***[0.16] [0.20] [0.16] [0.16] [0.12] [0.11] [0.10] [0.10]
Spices 0.02 0.01 0.02 0.02 0.11 0.05 0.09 0.03[0.13] [0.12] [0.12] [0.12] [0.13] [0.15] [0.11] [0.12]
Raw textile -0.28 -0.28 -0.26 -0.28 -0.26 -0.38 -0.32 -0.33[0.27] [0.26] [0.28] [0.28] [0.36] [0.36] [0.41] [0.40]
Fur, skin and leather -0.04 -0.04 -0.03 -0.03 0.03 0.01 0.05 0.06[0.17] [0.17] [0.18] [0.18] [0.21] [0.22] [0.22] [0.21]
Cloth -0.53*** -0.53*** -0.52** -0.53** -0.56** -0.69*** -0.73*** -0.72***[0.18] [0.19] [0.19] [0.19] [0.25] [0.25] [0.18] [0.18]
Geography and time YES YES YES YES YES YES YES YESInstitutions YES YES YES YES YES YESTrade geography YES YES YES YESPopulation YES YES
Observations 304 304 304 304 304 304 304 304City Clusters 26 26 26 26 26 26 26 26Century Clusters . . . 5 . . . .(Pseudo) R-squared 0.55 0.57 0.60 0.60 0.42 0.52 0.54 0.54Log (pseudo-)likelihood . . . . -96.0 -79.6 -75.1 -74.9Notes: ***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors clustered by city code (and by century for column(4)) in brackets. Columns (1)-(4) give OLS results for the choice of unit fees (vs value fees) for all products when the“matchmaking” mechanism was used. Column (1) controls for time and geographical coordinates, (2) incorporates theinstitutional control variables, column (3) the trade geographic variables, and column (4) population. Column (5)-(8)report marginal effects for corresponding probit regressions. Results for the product categories property, horses, andconstruction material in columns (5)-(8) are omitted: for property and horses, all observations were of matching withvalue fees, and for construction, all observations were with unit fees. Results for the regression in column (8), but withstandard errors clustered by century, can be found in the online appendix, in column (6) of Table 12.
21
the fee basis. Spices and similar goods such as coloring products fit this description, as do products
for the clothing manufacturing “industry”: raw textile inputs such as wool, linen, or fustian, (semi-
finished) cloth, and also furs, skin, and leather products. This might be surprising for some products
based on the pattern we have observed so far. Some of these products differed strongly in value and
quality, and we expect that buyers had fairly heterogeneous preferences. However, it appears that
towns followed two different strategies, which explains this observed use of different fee structures.
For the product category of spices, but also for cloth, fur, leather, and skins, we find two types of
brokerage fee structures. In cases where unit fees were used, we find long detailed lists for instance
of different spices, with a different unit fee assigned for each one. (One example is brokerage rules
from Frankfurt in 1373 – see Buecher (1915), p.250.) This reduced the heterogeneity within the
category. On the other hand, when value fees were applied, only a general expression for spices
was mentioned, with a fixed percentage applying to all of them. (An example is rules from Ofen in
1403 – see Michney and Lichner (1845), p. 74; or Bruinswick in 1320, see Hanselmann and Mack
(1900), vol. 2, pp. 516-7.)25,26
Most control variables do not show any significant effect when comparing fee structures in the
larger sample. In the smaller sample of matchmaking, we find some evidence of an effect of the
variables previously related to economic activities: cities with stronger overall economic activity
seemed somewhat more prone to use value rather than unit fees. Finally, product categories have
no impact on other details of the brokerage design.27
25In some cities, particularly during the seventeenth century, we find a combination of both systems at work: longlists of products with unit fees, plus a group of products with percentage fees. For Amsterdam, see Noordkerk (1748),pp.1060-1063; for Hamburg see Beukemann (1912), pp. 542-561. In cases where we found both types of fees in oneobservation, we coded this as value fees in the empirical analysis.
26Harder to explain is the frequent use of value fees for raw textiles, as these products are rather homogeneous.Indeed, when these products are traded by unit fees, we rarely find longer lists of products with different fees as inthe case of spices: we find one unit fee for a sack of wool (Cologne 1406, in Stein II, 1893-5, pp. 113f.) or a bale offustian (Middleburg 1405, in: Pols 1888, p.597). Nevertheless, we also frequently observe regulations specifying valuefees. In the case of regulations from Bruinswick, we find a switch from unit fees (in 1320) to value fees (in 1433).
27The output for the control variables can be found in the Online Appendix. In addition, a comparison betweenthe matching design and the second most frequent design (matching without a private business constraint) can befound. Introducing a private business constraint into the brokerage regulations is strongly determined by the politicalvariables: towns with strong governments such as a bishop seat or a local parliament (Free and Imperial cities) weremore likely to implement such a regulation. Both can be explained by a stronger legislature and enforcement of thesemarket regulations. This is in line with what we found before related to the general implementation of brokerageregulations.
22
3.4 Economic and Political Motivation
Many towns explained their motivations in regulating brokerage with short policy statements at
the beginning of the regulations. Most of these policy statements were based around one of five
broad goals: to promote and facilitate trade; to reduce the harm done to trade; to benefit the town;
to benefit the citizens; to ensure equal treatment of all merchants (locals and foreigners, rich and
poor) by brokers.
Typically, none of these explanations came with further detail. However, the goals are clearly in
line with what has been described in the literature on medieval town policy and trade (Heckscher
1933). Cities had two basic, partly opposing interests. On one hand, they had an interest in
ensuring a basic supply of consumption goods to the local population, and in providing the local
craft industry with raw and semi-finished input goods, at reasonable prices. On the other hand,
to guarantee such an inflow of products, towns had to offer foreign merchants attractive markets,
with favorable allocation mechanisms guaranteeing them a certain share of the surplus from trade;
otherwise merchants would go elsewhere. In addition, the equal treatment clearly reflects an en-
deavour to create open access order institutions in the sense of North et al. (2008). The frequency
of appearance of these policy statements is in line with the frequency of the combination of rules
appearing in the sample.28
4 Theory Model
Given the stated policy goals accompanying cities’ regulations, it’s natural to ask whether the
regulations chosen indeed created the correct incentives for broker behavior and trader participation,
and ultimately if this allocation process led to welfare-improving effects. In this section, we present
a simple theory model with three goals in mind: (i) to understand the incentives facing a broker, and
the effect that they have on the outcomes reached; (ii) to understand buyers’ and sellers’ incentives
to use the broker or try to trade on their own; and (iii) to build intuition for when brokerage creates
the most value; and to see how each of these vary with the details of the environment, and with the
choice of how brokerage is implemented. What we find helps to rationalize the empirical patterns
described above. In particular, we find theoretically that when products are very heterogeneous
28Table 14 in the Online Appendix shows how often each of these explanations appeared in the regulations.
23
in quality, value fees always lead to a more efficient outcome than unit fees, and encourage more
traders to use the broker; while when products are fairly homogeneous in quality and sellers are
therefore differentiated more by their costs, unit fees lead to a more efficient outcome than value
fees and encourage more traders to use the broker.
We begin by analyzing the broker’s choice of which buyers to pair with which sellers, given a
fixed set of traders seeking to trade through the broker; after that, we consider the incentives of
buyers and sellers to trade through the broker instead of on their own. The model is of course quite
stylized, and does not aim to capture the full complexity of the situation; but it is rich enough to
show how the broker’s incentives differ with the nature of the product and the basis of his fees.
4.1 Model
There is a fixed set of sellers s ∈ S = {1, 2, . . . , S}, each with a single good to sell. The sellers
potentially vary in both the quality of the good they are selling qs, and their cost cs, which we can
think of either as the cost of supplying the good or the seller’s value from keeping it (presumably
to sell later). We assume quality and cost vary in the same direction – a seller with higher quality
will also have higher cost.29
There is a fixed set of buyers b ∈ B = {1, 2, . . . , B}, each wishing to buy a single good. Each
buyer has a type vb indicating his valuation for goods of a particular quality level. Specifically, if
buyer b buys from seller s at price p, then the buyer and seller receive payoffs
vbqs − p and p− δ(p)− cs
respectively, where δ(p) is the commission paid to the broker (which might be a function of p).
Fixing the broker’s fee schedule δ(p), we will say a buyer b and seller s are compatible if there is
some p such that vbqs − p and p − δ(p) − cs are both strictly positive, i.e., if the buyer and seller
can both receive strictly positive payoffs by trading with each other after paying the broker’s fee.
Define a buyer as serious if he’s compatible with at least one seller, and likewise define a seller as
serious if he’s compatible with at least one buyer.
29This means our model has one-dimensional types – that is, we could equivalently assume each seller has a typeθs ∈ R, and offers a good with quality qs = q(θs) and cost cs = c(θs), where q(·) and c(·) are both weakly increasing.We discuss in the Online Appendix what happens if quality and cost vary in opposite directions. We do not haveresults on what would happen if quality and cost vary independently.
24
We consider a full-information environment, where the broker knows the types of each buyer
and each seller. (We will discuss later the incentives for traders to mislead the broker.) We consider
two cases: one, where the broker’s commission is proportional to the number of trades that occur
(“unit fees”); and the other, where the broker’s commission on each trade is proportional to the
price paid (“percentage fees”). Trade occurs in the following way:
1. The set of buyers and sellers attempting to trade through the broker is fixed, as is the broker’s
commission structure and fee level.
2. The broker pairs up buyers and sellers in whichever way maximizes his fees, subject to the
constraint that he can only pair up traders who are compatible.
3. A buyer and seller who are paired up together trade at the Nash bargaining price,30 which is
the solution to the problem
maxp
(vbqs − p)φ (p− δ(p)− cs)1−φ
We also assume there is a small probability that the buyer and seller fail to agree on a price,
and therefore fail to trade. We assume this probability is decreasing in the gains from trade,
and we take the limit as this probability becomes both vanishingly small and highly convex.
Specifically, we assume trade happens with probability
1− e−K(vbqs−cs−δ)
and take the limit as K →∞.31
Note that we do not consider each trader’s “outside option” when they negotiate over price.
30This outcome with φ = 12
was proposed by Nash (1950, 1953), who showed it uniquely follows from a particularset of axioms. Binmore, Rubinstein and Wolinsky (1986) discuss the use of this model as a “reduced form” formore complex dynamic bargaining games. Here, we can think of φ as representing the two sides’ relative bargainingstrengths; this can be thought of as capturing factors excluded from our model, such as the relative number of buyersand sellers or the likely time until the arrival of other merchants selling similar goods.
31This is done primarily to break the broker’s indifference among many potential pairings of buyers and sellers whenhis commission is the same for every trade; minimizing the expected number of trades which will fail to materialize,even when this probability is vanishingly small, serves as a “tiebreaker” in the broker’s preferences. The exactfunctional form is not important, this is simply one that works. Theorem 1 would still hold if we instead assumedthat among matchings with the same number of trades, the broker receiving unit fees would select the most efficientmatching. Part 2 of Theorem 1 would hold under any tiebreaking rule, and Part 1 of Theorem 1 would hold for anytiebreaking rule in the special case of uniform quality (qs = qs′ for any two sellers).
25
We imagine that while the eventual price to be negotiated is predictable, haggling is still a time-
consuming process, and that once a buyer and seller are paired up and begin to negotiate, neither
will have time to find another trading partner in the event they fail to agree. Thus, each can either
trade with this partner or not trade, and outside options are therefore irrelevant. Similarly, we do
not constrain the broker to select a stable matching – that is, to avoid matching a buyer b and
a seller s to other partners when they would prefer to trade with each other – but allow him to
select any matching of buyers to sellers subject to the constraint that any buyer-seller pair must
be compatible in order to trade. (We will consider stability later.) Informally, we imagine that if
it were easy for buyers and sellers who wanted to trade with each other to find each other, there
would be little need for the broker to begin with.
While we allow sellers to differ in both quality and cost, we will focus on the cases where one
type of variation dominates the other in terms of buyer preferences:
Definition 1. We will say that variation in costs dominates variation in quality (or “costs
dominate”) if for every serious buyer b and any two serious sellers s and s′,
cs < cs′ −→ vbqs − cs > vbqs′ − cs′
Conversely, we’ll say that variation in quality dominates variation in costs (or “quality
dominates”) if for every serious buyer b and any two serious sellers s and s′,
qs > qs′ −→ vbqs − cs > vbqs′ − cs′
That is, costs dominate if every relevant buyer would generate greater surplus trading with the
lower-cost of two sellers – that is, if every buyer agrees that the lower-cost product, even though it
is lower quality, is still a better value. Quality dominates if the reverse is true – even though the
higher-quality product is higher cost, every relevant buyer agrees it’s a better value. A special case
of costs-dominate is when all sellers offer identical quality, and a special case of quality-dominates
is when all sellers have identical costs.
We make two additional modeling assumptions. First, we assume no two traders are identical,
26
i.e., vb 6= vb′ for any two buyers b′ 6= b, and (qs, cs) 6= (qs′ , cs′) for any two sellers s′ 6= s.32 And
second, we assume that if a buyer b and a seller s are compatible under one type of fee (unit or
percentage), they are also compatible under the other type of fee. This is because we want to focus
on the way the broker’s incentives are shaped by the type of commission he receives, and this is
easiest to see when we can compare his choice under the two fee structures when facing the same
“choice set” (the same set of potential matchings). In practice, this holds if brokerage fees are small
relative to the gains from trade among compatible pairs.
Note that when we calculate social surplus, the broker’s fee is lost to the buyer and seller but
gained by the broker, so the social surplus generated by a trade is vbqs − cs, not vbqs − cs − δ(p).
4.2 How Does Brokerage Perform (Efficiency)
We begin by characterizing the outcomes that brokerage will lead to. We find that the way in which
sellers (or products) differ from each other affects which commission structure leads to higher overall
surplus.
Theorem 1. Fix the set of buyers and sellers, and assume brokerage fees are small.
1. If variation in costs dominates variation in quality, then unit fees always lead to a weakly
more efficient outcome than percentage fees.
2. If variation in quality dominates variation in costs, then percentage fees always lead to a
weakly more efficient outcome than unit fees.
The theorem is proved in the appendix, but much of the intuition can be gained from two simple
examples. First, consider a case where costs dominate, because all sellers offer identical quality.
Example 1. There are two buyers, with types (v1, v2) = (10, 6), and three sellers, with costs
(c1, c2, c3) = (4, 7, 9) and identical quality q1 = q2 = q3 = 1. Unit fees are less than 1, and
percentage fees less than 10%.
In this setting, total surplus is maximized if buyer 1 buys from seller 1 and nobody else trades;
this generates surplus of 10 − 4 = 6. A broker maximizing unit fees would pair buyer 1 with
32Given our other assumptions, this ensures a unique matching of buyers to sellers that maximizes total surplus,and a unique matching that maximizes expected commissions under either unit or percentage fees.
27
seller 2, and buyer 2 with seller 1, to get two trades instead of one; this would generate surplus of
(10 − 7) + (6 − 4) = 5. A broker maximizing percentage fees, however, would pair buyer 1 with
seller 3, and buyer 2 with seller 1. (That is, he would inflate the price the high-value buyer pays,
by pairing him with the highest-cost seller.) This would generate surplus of (10− 9) + (6− 4) = 3.
In this example, either type of fees leads to more than the efficient number of trades, but
percentage fees leads to a further inefficiency: the broker increases his commission by pairing buyer
1 with the highest-cost seller instead of the second-highest-cost, leading him to pay a higher price.
In fact, the features of Example 1 generalize. The proof of Theorem 1 part 1 is based on two
facts which hold whenever variation in costs dominates: (i) unit and percentage fees always lead
to the same number of trades, and (ii) unit fees always lead the broker to select the most efficient
matching with this number of trades, while percentage fees may not.
Next, we consider an example where quality dominates, because all sellers have identical costs.
Here, there is a potential for a different type of inefficiency – a mismatch between seller quality
and buyer willingness-to-pay-for-quality – which turns out to be far more severe under unit fees.
Example 2. There are four buyers, with types (v1, v2, v3, v4) = (8, 5, 3, 2), and four sellers, with
quality (q1, q2, q3, q4) = (6, 4, 3, 2) and identical costs c1 = c2 = c3 = c4 = 10. For simplicity, we
assume φ ≈ 0 – sellers have all the bargaining power, so trades all occur at the buyer’s valuation
vbqs. Again, unit fees are less than 1, and percentage fees less than 10%.
In this environment, total surplus is maximized if buyer 1 gets the highest-quality object, buyer
2 gets the second-highest quality object, and the other two objects don’t get traded; this generates
surplus of (8× 6− 10) + (5× 4− 10) = 48.
Percentage fees are maximized if buyer 1 gets the highest-quality object, buyer 2 gets the third -
highest quality object, and buyer 3 gets the second-highest-quality object. (This generates a third
transaction, and prices of 48, 15 and 12 rather than 48 and 20, so the broker earns a commission on 7
more dollars of sales.) Total surplus is lower than before, however, at (48−10)+(15−10)+(12−10) =
45. (Selling good 3 to buyer 2 and good 2 to buyer 3 generates higher commissions than just selling
good 2 to buyer 2, but less surplus.)
Unit fees are maximized if all four objects trade; this is only possible if buyer 1 gets good 4,
buyer 2 gets good 3, buyer 3 gets good 2, and buyer 4 gets good 1. This is hugely inefficient: the
28
higher-quality goods are largely “wasted” by going to lower-value buyers, and total surplus is just
(16− 10) + (15− 10) + (12− 10) + (12− 10) = 15.
When quality dominates, we can show that percentage fees may lead to more than the efficient
number of trades, but always lead to the most efficient matching with that many trades; while unit
fees may lead to even more trades, and may not lead to the most efficient matching with that many.
Overall, then, the result is quite sharp: when sellers offer similar-quality products and differ
mostly in their costs, unit fees lead to a more efficient outcome; and when sellers instead offer
products with widely-varying quality but have similar costs, percentage fees lead to a more efficient
outcome. Note also that these results hold for every realization of buyer and seller types, not just
in expectation over certain probability distributions of types.
4.3 Incentives to Use the Broker (Stability)
As we noted earlier, there were rare instances in which merchants were forced to trade through
brokers, but this was the exception rather than the rule; for the most part, merchants could choose
whether to trade through a broker or on their own. Thus, we want to consider the incentives for
traders to elect to trade through the broker. Of course, if the broker’s fees are large, traders have
an incentive to search on their own even if they would be perfectly happy with the broker-selected
match, purely to avoid paying the commission. We abstract away from this by assuming the fee
is small, and instead focus on traders’ incentives to search on their own in the hope of improving
their match relative to the partner the broker would pair them with.
We do this by considering the stability of the matching selected by the broker.33 A matching
M is a mapping from the set of traders to itself, where M(b) = s and M(s) = b if buyer b and seller
s are paired together (and a trader who does not trade is mapped to himself). If we let µ(b, s) be
the payoff buyer b gets from buying from seller s at the Nash price, and ν(s, b) the payoff seller s
gets from selling to buyer b, a matching M is said to be stable if there is no pair of traders (b, s)
such that
µ(b, s) > µ(b,M(b)) and ν(s, b) > ν(s,M(s))
that is, if there is no buyer-seller pair who are not matched to each other, who would both happily
33There is an enormous literature considering stability in matching markets. See Roth and Sotomayor (1990) fora review of the early literature, and Abdulkadiroglu and Sonmez (2013) for a more recent survey.
29
abandon their partner under M to instead trade with each other. If a broker’s incentives led him
to select a stable matching, this would mean that no group of traders could all benefit, even with
full information and no frictions, from abandoning the broker to trade among themselves. We can
interpret this as meaning participation in brokerage would likely be high.34
(In contrast with our approach, Bloch and Ryder (2000) study equilibrium in a similar but
distinct model when each trader chooses for himself whether to try to trade through the broker or
on their own. Bloch and Ryder, however, assume the broker is constrained to select the (typically
unique) stable matching, so that the only reason to forego the broker would be to avoid paying
his fee; and the game they study has a large multiplicity of equilibria. In contrast, we assume the
broker is free to choose a matching which is not stable, and focus on traders’ incentives to search
on their own in the hopes of improving one’s match.)
Theorem 3 in the Online Appendix shows one special case where percentage fees lead the broker
to select the (unique) stable matching. Outside of this particular special case, though, stability in
our setting is generally not guaranteed. In Examples 1 and 2 earlier, for example, the matching
chosen by the broker under either fee structure was not stable. Thus, stability of the chosen match
appears to be too high a bar for us to use. However, we can also consider how close to stable an
outcome each fee structure leads to. For a given value ε ≥ 0, a matching M is called ε-stable if
there is no buyer-seller pair (b, s) such that
µ(b, s) > µ(b,M(b)) + ε and ν(s, b) > ν(s,M(s)) + ε
that is, if no buyer and seller could both gain more than ε by instead trading with each other. (We
can therefore think of ε as the level of fixed cost of private search which would suffice to make all
traders trade through the broker, or more vaguely as a measure of how strong an incentive buyers
face to try to trade without the broker.) We’ll define a matching M ′ to be weakly more stable than
another matching M if for any value of ε, if M is ε-stable then M ′ must be as well. And we can
give results on ε-stability which mirror our results on efficiency:
34Games where each trader chooses whether to trade through the broker or on his own always have multipleequilibria, since if all the buyers seek to trade through the broker, all the sellers will want to as well, and if all thebuyers search on their own, all the sellers will want to. Stability can be thought of as meaning that everyone tradingthrough the broker would still be an equilibrium, even if traders had full information about each others’ types andno frictions in finding each other.
30
Theorem 2. Fix the set of buyers and sellers, and assume brokerage fees are small.
1. If variation in costs dominates variation in quality, then unit fees always lead to a weakly
more stable outcome than percentage fees.
2. If variation in quality dominates variation in costs, then percentage fees always lead to a
weakly more stable outcome than unit fees.
If unit fees (when costs dominate) or percentage fees (when quality dominates) lead to a more
stable matching, this suggests the broker should attract more traders, as traders have less incentive
to bypass him if there’s a cost to searching and trading on their own. Thus, our results on stability
seem to reinforce our results on efficiency. When costs dominate, unit fees lead to a more efficient
outcome than percentage fees for a given set of traders trading through the broker, and give more
of an incentive for traders to use the broker; when quality dominates, it is percentage fees that lead
to both a more efficient outcome and greater incentives to trade through the broker.35
(In the appendix, we give a modified version of Examples 1 and 2 which illustrate why the
stability results hold.)
4.4 When Is Brokerage Valuable
Finally, we consider the question of whether brokerage increases total surplus, and when it is most
valuable. To address this question, we imagine that in the absence of brokerage, buyers and sellers
would pair up randomly, and attempt to trade with whatever trading partner they ended up with.
If there were more buyers than sellers, each seller would find a buyer, but each buyer would have
a probability less than one of finding a seller; vice versa if there were more sellers than buyers.
Consistent with our assumptions about trade within brokerage, we assume that traders would only
have one opportunity to trade – they pair up with someone (or don’t), but by the time they figure
out whether they are compatible and attempt to work out a deal, no other opportunities exist for
trade.
Theorem 1 above establishes conditions under which one type of fees (either unit or percentage)
is certain to be more efficient than the other, but it does not guarantee that either will lead to
35Note, however, that the stability results do not automatically strengthen the efficiency results. For example,when quality dominates, if the highest-value buyer and highest-quality seller opt out of brokerage to trade directlywith each other instead of matching with other traders through the broker, this increases total surplus, provided thecost to them of finding each other is not too high.
31
the most efficient possible matching, or even that either one will be more efficient than random
matching on average. In fact, it is easy to generate examples of small markets in which random
matching would outperform brokerage with either types of fees. To see when brokerage tends to
give the largest benefit relative to random matching – and therefore the conditions under which we
would most expect to see brokerage occur – we have run simulations on markets of different sizes
and configurations. Details are given in the appendix. The general takeaway from the simulations
is as follows:
1. In balanced markets – markets with equal numbers of buyers and sellers – brokerage offers
little to no improvement over random matching. For example, in the market we simulated
with 5 buyers and 5 sellers with costs dominating, random matching on average achieves
73% of the total surplus theoretically available. Brokerage with unit fees does slightly better,
achieving 79% of the theoretical maximum, while brokerage with percentage fees does slightly
worse, achieving 66% of the theoretical maximum. With 10 buyers and 10 sellers, brokerage
with either type of fees is less efficient than random matching; results are similar when quality
dominates.
2. In buyer-heavy markets – markets with more buyers than sellers – brokerage with either unit
or percentage fees gives a large efficiency gain over random matching. Regardless of whether
costs or quality dominates, unit and percentage fees perform almost equally well, and both
get fairly close to the maximal level of total surplus theoretically possible. For example, in the
market we simulated with 30 buyers and 10 sellers when costs dominate, random matching
on average generates only 46% of the maximal available surplus, while brokerage under either
unit or percentage fees generates 92% of the available surplus; results were again similar when
quality dominates.
3. In seller-heavy markets – markets with more sellers than buyers – brokerage offers a large
efficiency gain, but only when the right type of fees are used. For example, with 10 buyers and
30 sellers when costs dominate, random matching again achieves about 46% of the available
surplus; matching with unit fees achieves 92%, but matching with percentage fees achieves
only 11% of the available surplus. In the same size market when quality dominates, random
32
matching achieves 49% of the available surplus; matching with unit fees achieves only 23%,
while matching with percentage fees achieves 96% of the available surplus.
Thus, based on our simulations, brokerage seems to be most desireable in unbalanced markets;
when buyers outnumber sellers, unit fees and percentage fees both perform well, but when sellers
outnumber buyers, only the type of fees favored by Theorem 1 (unit fees when costs dominate,
percentage fees when quality dominates) perform well. In particular, the patterns documented
in the data – the use of brokers as matchmakers, and the more predominant use of unit fees for
relatively homogeneous goods and percentage fees for relatively heterogeneous ones – would seem to
work quite well in either sort of unbalanced markets, whether it was buyers or sellers who were more
plentiful. (We might expect such imbalances to be common in emerging and developing markets,
where temporary product shortages and price volatility were common.)
4.5 Discussion of Modeling Choices
In our model, we focus on efficiency – maximizing total surplus – rather than, say, maximizing the
number of trades or the surplus earned by buyers. This is because brokerage can be thought of as a
two-sided platform, needing to attract both buyers and sellers. As discussed earlier, town officials
would have been most interested in the welfare of their own citizens, not foreign merchants; but they
also needed to attract merchants to the town. Under our assumption of Nash bargaining, buyers
and sellers split the total gains from trade in a fixed ratio, so maximizing surplus (by matching the
right buyers to the right sellers) would maximize the payoff to both sides of the market.
In fact, the need to attract sellers to a town could easily be part of the explanation for why
brokerage appealed – rather than simply passing regulations to allow buyers to buy on more and
more favorable terms at the expense of sellers, well-designed brokerage could potentially increase
the gains to both sides of the market, encouraging sellers to come while still preserving the surplus
of the local buyers. This is especially true for towns on the same trade route, who competed with
one another for sellers’ time. It’s particularly striking in the data how brokerage would appear
in many places along the same routes. Once one town instituted brokerage, other towns which
were visited by the same sellers (or competed for them) very often also chose to pass similar rules,
suggesting the rules worked in creating a market which was beneficial to both sides.
33
We’ve chosen not to model entry by brokers, nor competition among brokers. (In contrast,
papers on middlemen – traders who buy only to resell – like Rubinstein and Wolinsky (1987)
and Nosal, Wong and Wright (2015) endogenize both the number of traders choosing to act as
middlemen, and the prices at which they buy and sell.) For our application, this seems unnecessary.
Most towns restricted who could act as a broker, so there was certainly not free entry. Many towns
had only a single broker for a given product, and when there were more than one, they were often
forced to share their commissions, so they were not in competition with each other; and the level
of commissions was set by town officials. Thus, it seems reasonable to assume a single broker
operating at an exogenously-chosen commission level.
We also do not model the prohibition on brokers buying and selling on their own account (the
private business constraint). This can be thought of as the choice to use brokers (matchmakers)
rather than middlemen. Our intuition from papers on middlemen (such as the two cited above) is
that since they bear the risk of being “stuck” with unsold goods they don’t value, they must earn
a much greater profit to compensate for that risk, relative to brokers who do not. Especially if
cities were competing to attract sellers, choosing a mechanism under which the intermediary earns
a more modest fee, and therefore leaves more of the gains from trade to the buyers and sellers,
would seem beneficial.
Finally, we simplified our model by assuming full information – that the broker knows both
buyers’ and sellers’ preferences, and does not have to concern himself with learning them.36 If
buyers and sellers had private information about their types, they might well have incentives to
misrepresent them to the broker, in order to get matched to a different trader or trade at a more
favorable price. This incentive is always present, but we would expect it to be smallest when “good”
traders are being rewarded with “good” matches, which tends to happen more with percentage
fees, especially in the quality-dominates case. Private information would also be another important
reason for the private business constraint – traders would presumably be much more willing to
reveal their true type when the broker was only going to use it to match them to a trading partner,
as opposed to being a middleman who was seeking to buy from them (or sell to them) at the most
advantageous price.
36In a different model in earlier versions of this paper, we assumed the broker knew buyers’ preferences but neededto elicit sellers’, and got qualitatively fairly similar results.
34
5 Conclusion
This paper studies the implementation of allocation mechanisms in pre-industrial economies. We
examine brokerage as a market-clearing institution implemented by town officials in late medieval
and early modern merchant cities. We show that cities implemented brokerage in environments
associated with stronger economic activity. In addition, we show that the specific brokerage mecha-
nisms chosen were sensitive to the product markets where they were applied, and that the patterns
in product-specific choices are broadly consistent with the welfare-maximizing ones.
To achieve these results, we created a comprehensive source analysis covering brokerage statutes
from 70 towns north of the Alps in Central and Western Europe from the middle of the 13th century
until the end of the 17th century, based on an investigation of 231 cities. We identified brokerage
as a centralized matchmaking institution implemented by towns to promote and support trade, to
create welfare for the town and their citizens, and to give equal opportunities to the citizens and
foreign merchants. We showed that the implementation of brokerage regulations correlates with
variables related to economic activity identified in the literature, and that variations in the dominant
brokerage matching design relate to different product categories with different characteristics. We
introduced a theoretical matching model, and showed that the empirical pattern found in the data
– price-based brokerage fees for products which are more heterogeneous in quality, volume-based
fees for products which are more homogeneous – are consistent with the welfare-maximizing choice.
Finally, we used simulations to show that brokerage is most valuable in unbalanced markets, and
that the choice of the appropriate type of fees is most crucial when sellers outnumber buyers.
These results show the important role of market design for economic development. Our findings
highlight the active role of merchant cities in pre-industrial Europe in creating markets and solving
incentive and allocation problems. They support the argument of an institutional change starting
during this period which created large benefits for society. In particular, we identify centralized
allocation mechanisms as open access order institutions (North et al. 2006) and as market policy
instruments that have an impact on the welfare of both sides of the market (Heckscher 1933).
These results open up new questions for a future research agenda. This paper covered one
specific market-clearing mechanism out of a larger set of medieval market rules. Other rules related
to market location and the timing of buying and selling goods. It would be of considerable interest
35
to examine the interplay of brokerage and other market mechanisms. Another question relates to
the evolution, adaptation, and learning behavior of market rules over time. Finally, it would be
interesting to link market making activities more explicitly to city growth and understand causation,
with the ultimate goal to understand divergence in growth paths in different cities and regions.
36
Appendix
A.1 Description of Explanatory Variables
VARIABLE DESCRIPTION
Free Imperial Towns A binary variable taking the value 1 if the town is either an “imperial” or “free” city
and 0 otherwise. Imperial cities were directly ruled by a local consulate representing
(some of) their citizens, but were formally under the legal protection of the German
Kaiser (who had normally only limited influence on the political decision making
of a town). The same holds for free cities, but they liberated themselves from the
regimen of a bishop. (See Heining 1983, Johannek 2000.) For the identification of
Free and Imperial cities, see Johanek (2000).
Bishop A binary variable taking the value 1 if a town is ruled by a bishop and 0 otherwise.
Bishop cities are for instance documented in Bautier et al. (1977-1999).
Territorial Towns A binary variable taking the value 1 for towns which were neither free, imperial, nor
bishop towns. All other towns were controlled by local (territorial) dukes (see Isen-
mann 1988). Note this variable is not included in the regression since the political
town characteristics are already sufficiently specified.
Hanseatic A binary variable which is 1 if the city joined the Hanseatic League and 0 otherwise.
The information on the participation can be found in Dollinger (1966).
University A binary variable which is 1 if a city hosts a university and 0 otherwise. A history of
the foundation and evolution of university towns can be found in de Ridder-Symoens
and Ruegg (1992-2011).
Roman A binary variable which is 1 if a town was founded by Romans and 0 otherwise. The
information about Roman towns can be found in Putzger (1956) and Bedon (2001).
Water A binary variable which is 1 if a town has a navigable port and 0 otherwise. The
information of navigable ports (river or sea) can be found in Putzger (1956), p. 70.
Sea A binary variable which is 1 if a town has a port with access to the sea and 0
otherwise. This information can be found in Putzger (1956), p.70.
Year This variable identifies the specific year a brokerage regulation has been dated. In
the empirical analysis on the existence of brokerage, it assigns the year to the 50-year
interval in which it is located.
Longitude This variable describes the longitude a city is located. (see google maps)
Latitude This variable describes the latitude a city is located. (see google maps)
Log(Population) The logarithm of the city population size, based on data collected by Bairoch (1988),
interpolated when necessary
Population Quintile These are five binary variables dividing the sample observations in five equally
weighted quintiles, ordered by the city population size (quintile 1 being the small-
est population size). Each variable is 1 if the observation can be assigned to the
respective quintile and 0 otherwise.
37
A.2 Examples for understanding Theorem 2
Example 1′. Example 1 had two buyers, with (v1, v2) = (10, 6), and three sellers, with (c1, c2, c3) =
(4, 7, 9) and q1 = q2 = q3 = 1. Suppose φ = 12 and fees are small.
Under unit fees, the broker matches buyer 1 with seller 2 and buyer 2 with seller 1. Under
percentage fees, he matches buyer 1 with buyer 3 and buyer 2 with seller 2.
• Under unit fees, buyer 1 gets surplus 12(10− 7) = 3
2 and seller 1 gets surplus 12(6− 4) = 1. If
buyer 1 and seller 1 instead matched together, they would each get surplus of 12(10− 4) = 3.
Thus, the matching under unit fees is ε-stable for all ε > 3− 32 = 3
2 .
• Under percentage fees, buyer 1 gets surplus 12(10 − 9) = 1
2 and seller 1 still gets surplus 1.
If they matched together instead, they would each get 3. So the matching under unit fees is
ε-stable for all ε > 3− 1 = 2.
Thus, the matching under unit fees is more stable than the matching under percentage fees.
Example 2′. Example 2 had four buyers, with (v1, v2, v3, v4) = (8, 5, 3, 2), and four sellers, with
(q1, q2, q3, q4) = (6, 4, 3, 2) and c1 = c2 = c3 = c4 = 10. Suppose now that φ = 12 , and fees are low.
Under unit fees, the broker matches buyer 1 to seller 4, 2 to 3, 3 to 2, and 4 to 1. Under
percentage fees, he matches buyer 1 to seller 1, buyer 2 to seller 3, and buyer 3 to seller 2.
• Under unit fees, buyer 1 gets surplus 12(8·2−10) = 3 and seller 1 gets surplus 1
2(2·6−10) = 1.
If they matched together instead, they would each get 12(8 · 6− 10) = 19. Thus, the matching
under unit fees is only ε-stable for ε > 16.
• Under percentage fees, buyer 1 and seller 1 are already matched together, so they have no
need to consider a change. Buyer 2 is getting surplus 12(5 · 3− 10) = 5
2 and seller 2 is getting
surplus 12(3 · 4− 10) = 1; if they matched together instead, they’d each get 1
2(5 · 4− 10) = 5.
So the matching under percentage fees is ε-stable for ε > 52 .
This time, the matching under percentage fees is more stable than the matching under unit fees.
A.3 Simulation Results on Value of Brokerage
As noted in the text, we used numerical simulation to compare the outcome with a broker to the
outcome under random matching. The simulations were performed in NetLogo.
Our baseline specification for the costs-dominate case had buyer types vb drawn from the uniform
distribution on the interval [5, 10], constant quality qs = 1 for all sellers, and seller costs cs drawn
from the uniform distribution on the interval [5, 10]. Our baseline specification for the quality-
dominates case had vb drawn uniformly from [5, 10], qs drawn uniformly from [5, 10], and constant
costs cs = 50. Throughout, we assumed φ = 12 (the buyer and seller split the gains from trade
evenly) and the broker’s fees were small enough to ignore.
38
For each specification, for markets with different numbers of buyers and sellers, we randomly
generated 100, 000 copies of the economy (sets of buyer and seller preferences); calculated the
expected surplus realized under random matching based on 1,000 randomly-generated matchings
of buyers to sellers; and calculated the surplus realized in the matchings that maximize unit fees,
percentage fees, and total surplus. We then averaged over the 100,000 simulations to get an
estimate of expected surplus for each matching process (random, unit fees, percentage fees, efficient
matching).
In the tables below, for each size market and each specification of the model, we show the
percentage increase in total surplus from switching from random matching to each of the others.
To begin, we look at “balanced markets” – markets with equal numbers of buyers and sellers.
The first row of Table 7 shows that in a market with 2 buyers and 2 sellers, in the baseline costs-
dominate specification, switching from random matching to brokerage with unit fees would increase
total surplus by 10%; switching from random matching to brokerage with percentage fees would
leave total surplus unchanged; and switching from random matching to the most efficient possible
matching would increase total surplus by 20%. Our takeaway from this table is that the benefit
of brokerage (relative to random matching) is rather marginal, and often negative, in balanced
markets.
Table 7: Baseline specification, balanced marketsCosts Dominate Quality Dominates
Number of Number of Gain from Gain from Gain from Gain from Gain from Gain fromBuyers Sellers Unit Fees Pct Fees Efficient Unit Fees Pct Fees Efficient
2 2 10% 0% 20% -3% 5% 16%3 3 11% -3% 29% -6% 4% 24%5 5 8% -9% 36% -12% 1% 30%10 10 -4% -22% 43% -23% -9% 36%20 20 -19% -37% 46% -34% -21% 40%50 50 -41% -55% 49% -46% -34% 41%
Next, we consider markets where buyers outnumber sellers. Table 8 shows that in these markets,
the gains from brokerage are quite large, and nearly as large as moving to the efficient matching,
regardless of which type of fees are used; and the gains from brokerage are larger as the market
gets more unbalanced.
Next, we consider markets where sellers outnumber buyers. Table 9 shows that the gains from
brokerage are again large, but now depend on using the “right kind of fees” – unit fees when
costs dominate, percentage fees when quality dominates. While the gains from brokerage with the
appropriate fee type are large, and get larger as the market gets more unbalanced, brokerage with
the “wrong kind of fees” is worse than random matching (and gets even worse as the market gets
more unbalanced).
Finally, Tables 15 and 16 in the Online Appendix show results for different variations in the
distributions of buyer valuations, seller quality, or seller costs. These are intended primarily as
39
Table 8: Baseline specification, buyers outnumber sellersCosts Dominate Quality Dominates
Number of Number of Gain from Gain from Gain from Gain from Gain from Gain fromBuyers Sellers Unit Fees Pct Fees Efficient Unit Fees Pct Fees Efficient
4 2 63% 60% 72% 52% 55% 63%6 3 66% 63% 80% 53% 59% 70%10 5 66% 63% 88% 54% 62% 77%20 10 62% 61% 93% 53% 64% 82%40 20 58% 57% 97% 52% 65% 85%
6 2 94% 93% 100% 79% 83% 87%9 3 98% 97% 108% 82% 87% 93%15 5 101% 100% 114% 84% 91% 99%30 10 101% 101% 119% 86% 95% 103%60 20 101% 101% 122% 86% 97% 106%
10 2 128% 127% 131% 107% 109% 111%15 3 132% 132% 137% 110% 114% 116%25 5 136% 135% 142% 112% 117% 120%50 10 138% 138% 146% 114% 120% 124%100 20 139% 139% 148% 115% 122% 125%
Table 9: Baseline specification, sellers outnumber buyersCosts Dominate Quality Dominates
Number of Number of Gain from Gain from Gain from Gain from Gain from Gain fromBuyers Sellers Unit Fees Pct Fees Efficient Unit Fees Pct Fees Efficient
2 4 63% -15% 71% -12% 56% 63%3 6 66% -26% 80% -20% 59% 70%5 10 66% -42% 88% -31% 62% 77%10 20 62% -62% 94% -45% 64% 82%20 40 58% -77% 97% -54% 65% 85%
2 6 94% -30% 100% -23% 83% 87%3 9 98% -44% 108% -32% 87% 93%5 15 101% -60% 114% -42% 91% 99%10 30 101% -76% 119% -52% 95% 104%20 60 101% -87% 122% -59% 97% 106%
2 10 128% -52% 131% -35% 109% 111%3 15 132% -64% 137% -43% 113% 116%5 25 136% -76% 142% -50% 117% 120%10 50 138% -87% 146% -56% 120% 124%20 100 139% -93% 148% -59% 122% 125%
robustness checks. While the exact numbers of course change from specification to specification,
the results are qualitatively consistent with those above.
40
Figure 1: The Existence of Brokerage in Central/Western Europe between 1250 and 1700.
Dots depict all towns under investigation, i.e., all towns with a population of at least 5,000 at some point during the period of investigation
(1200-1700). Towns with brokerage regulations are marked in red and labeled; other towns (without brokerage) are marked in black.
Source: Based on own source material.
41
Figure 2: Matching Mechanisms Along the Rhine-Main-Meuse-Scheldt Area.
The left panel shows towns which implemented the matchmaking mechanism. The right panel shows towns which implemented the matchmaking
mechanism for wine, ca. 1350-1400. Source: Based on own source material.
42
References
1. Abdulkadiroglu, Atila, and Tayfun Sonmez, 2013. “Matching Markets: Theory and Practice.”Advances in Economics and Econometrics, 1, 3-47.
2. Acemoglou, Daron and James Robinson, 2012. Why Nations Fail: The Origins of Power,Prosperity, and Poverty. New York: Crown Publishing Group.
3. Acemoglou, Daron, Simon Johnson, and James Robinson, 2005. The Rise of Europe: AtlanticTrade, Institutional Change and Growth, American Economic Review, 95 (3), 546-579.
4. Acemoglou, Daron, Simon Johnson, and James Robinson, 2005b. Institutions as a Fundamen-tal Cause of Long-Run Growth, in: Philippe Aghion and Steven N. Durlauf (eds.), Handbookof Economic Growth, Volume 1, Part A, 385-472.
5. Bairoch, Paul, 1988. Cities and Economic Development, from the Dawn of History to thePresent, Chicago: University Press.
6. Bairoch, Paul, Jean Batou, and Piere Chevre, 1988. La Population des villes Europeennes,banque des donnees et analyse sommaire des resultats: 800-1850, Geneva: Droz.
7. Bautier, Robert Henri, and others, 1977-1999. Lexikon des Mittelalters, 9 Volumes. Muenchen:Artemis.
8. Beukemann, Friedrich, 1912. Die Geschichte des Hamburger Maklerrechts, Heidelberg: CarlBeyer.
9. Beyerbach, Johann Conrad, 1818. Sammlung der Verordnungen der Reichsstadt Frankfurt,Frankfurt a. M.
10. Binmore, Ken, Ariel Rubinstein, and Asher Wolinsky, 1986. The Nash Bargaining Solutionin Economic Modelling, RAND Journal of Economics 17 (2), 176-188.
11. Bochove, Christiaan van, 2013. Configuring Financial Markets in Pre-modern Europe, Jour-nal of Economic History, Volume 73 (1), 247-278.
12. Bochove, Christiaan van, Lars Boerner, and Daniel Quint, 2017. Anglo-Dutch PremiumAuctions 18th Century Amsterdam, working paper, available athttp://www.ssc.wisc.edu/˜dquint/papers/anglo-dutch.pdf
13. Boerner, Lars, 2016. Medieval Market Making: Brokerage Regulations in late Medieval andEarly Modern Central/Western Europe. LSE Economic History Working Paper.
14. Boerner, Lars, and John William Hatfield, 2017. The Design of Debt Clearing Markets:Clearinghouse Mechanisms in Pre-Industrial Europe, Journal of Political Economy, Volume125 (6), 1991-2037.
15. Boerner, Lars, and Albrecht Ritschl, 2009. The Economic History of Sovereignty: CommunalResponsibility, the Extended Family, and the Firm, Journal of Institutional and TheoreticalEconomics, Volume 165 (1), pp. 99-112.
16. Bosker, Maarten, Eltjo Buringh, and Jan Luiten van Zanden 2013. From Baghdad to London:Unraveling Urban Development in Europe, the Middle East, and North Africa, 8001800.Review of Economics and Statistics Volume 95(4), 1418-1437.
17. Buecher, Karl (ed.), 1915. Frankfurter Amts- und Zunfturkunden bis zum Jahre 1612, Frank-furt a. M.: Baer.
43
18. Cantoni, Davide, and Noam Yuchtmann, 2014. Medieval Universities, Legal Institutions, andthe Commercial Revolution, The Quarterly Journal of Economics 129, 823-887.
19. Dietz, Alexander, 1910-1925. Frankfurter Handelsgeschichte, 5 Volumes. Frankfurt a. M.:Hermann Minjon.
20. Dijkmann, Jessica, 2011. Shaping Medieval Markets. The Organisation of Commodity Mar-kets, C.1200- C. 1450. Leiden: Brill.
21. Dollinger, Philippe, 1966. Die Hanse, Stuttgart: Alfred Kroner Verlag.
22. Donna, Javier, and Jose-Antonio Espın-Sanchez, 2015. The Illiquidity of Water Markets:Efficient Institutions for Water Allocation in Southeastern Spain. Working paper.
23. Friedland, Klaus, 1991. Die Hanse. Stuttgart: Verlg W. Kohlhammer.
24. Gelderblom, Oscar, 2013. Cities of Commerce, The Institutional Foundation of InternationalTrade in the Low Countries 1250-1650, Princeton: Princeton University Press.
25. Gehrig, Thomas, 1993. Intermediation in Search Markets, Journal of Economics and Man-agement Strategy 2, 97-120.
26. Geny, Joseph, 1902. Elsasische Stadtrechte, I Schlettstadter Stadtrechte, Heidelberg: CarlWinter’s Universitatsbuchhandlung.
27. Gilliodts- van Severen, L., 1881. Les Anciens Reglements de la Corporation de Courtiers deBruges, La Flandre, Revue des monuments d’histoire et d’antiquities XII, 121-148, 219-260.
28. Greif, Avner, 1993. Contract Enforceability and Economic Institutions in Early Trade: TheMaghribi Traders’ Coalition, American Economic Review 83 (3), 525-548.
29. Greif, Avner, 1994. Cultural Beliefs and the Organization of Society: A Historical and The-oretical Reflection on Collectivist and Individualist Societies, Journal of Political Economy102 (5), 912-950.
30. Greif, Avner, 2006. Institutions and the Path to the Modern Economy, Lessons from MedievalTrade, Cambridge: University Press.
31. Hanselmann, Ludwig, and Heinrich Mack (eds.), 1900. Urkundenbuch der Stadt Braun-schweig II and III, Braunschweig.
32. Hibbert, A. B., 1967. The Economic Policies of Towns, in: M. M. Postan et al. (eds.), TheCambridge Econoic History of Europe, Volume III, 157-230.
33. Hoehlbaum K., K. Kunze, H. G., Van Rundstedt, and W. Stein, 1876-1939. Hanisches Urkun-denbuch, 11 vol.
34. Holtfrerich, Carl-Ludwig, 1999. Frankfurt as a Financial Centre: From Medieval Trade Fairto European Banking Centre. Munich: Verlag C.H. Beck.
35. Hunt, Edwin S. and James M. Murray, 1999. A History of Business in Medieval Europe,1200-1550, Cambridge: Cambridge University Press.
36. Irsliger, 2010. Der Rhein als Handelsstrasse im spaten Mittelalter, in: Franz J. Felten (ed.),Wirtschaft an Rhein und Mosel. Von den Romern bis ins 19. Jahrhundert, Stuttgart: FranzSteiner Verlag.
37. Isenmann, Eberhard, 1988. Die deutsche Stadt im Spatmittelalter, 1250-1500, Stuttgart:Verlag Eugen Ulmer.
44
38. Lopez, Robert S., 1976. The Commerical Revolution of the Middle Ages 950-1350, Cambridge:Cambridge University Press.
39. Matheus, Michael (ed.), 2004. Weinproduktion und Weinkonsum im Mittelalter, Stuttgart:Steiner Verlag.
40. McCusker, John J. and Cora Gravesteijn, 1991. The Beginnings of Commercial and FinancialJournalism, Amsterdam: NEHA.
41. Michney and Lichner (eds.), 1845. Das Ofener Stadrecht von 1244-1421, Pressburg.
42. Milgrom, Paul, 2004. Putting Auction Theory to Work. Cambridge: Cambridge UniversityPress.
43. Moltke, Siegfried, 1939. Die Geschichte der Leipziger Maklerschaft. Leipzig: Deichert.
44. Munro, John, 1994. The International Law Merchant and the Evolution of Negotiable Creditin Late-Medieval England and the Low Countries, in: John Munro (ed.): Textiles, Towns,and Trade: Essays in the Economic History of Late Medieval England and the Low Countries,Ashgate: Aldershot.
45. Nash, John, 1950. The Bargaining Problem, Econometrica 18(2), 155-162.
46. Nash, John, 1953. Two-Person Cooperative Games, Econometrica 21(1), 128-140.
47. Neeman, Zvika, and Nir Vulkan, 2010. Markets Versus Negotiations: the Predominance ofCentralized Markets, The B.E. Journal of Theoretical Economics 10 (1), 1-30.
48. Noordkerk, H., 1748. Handvesten van Amsterdam. 3 volumes, Amsterdam.
49. North, Douglass C., 1981. Structure and Change in Economic History. New York: Norton.
50. North, Douglass C., 1990. Institutions, Institutional Change, and Economic Performance.Cambridge: Cambridge University Press.
51. North, Douglass C., John J. Wallis, and Barry R Weingast 2009. Violence and Social Orders:A Conceptual Framework for Interpreting Recorded Human History. Cambridge: CambridgeUniversity Press.
52. Pols, M. S., 1888. Westfriesche Stadtrechten, ’S Gravenhage: Martinus Nijhoff.
53. Postan, M. M., 1987. The Trade of Medieval Europe, in: Postan et al. (eds.). The CambridgeEconomic History of Europe, Volume 2, Cambridge: Cambridge University Press, 168-305.
54. Richardson, Gary, 2004. Guilds, Laws, and Markets for Manufactured Merchandise in Late-Medieval England, Explorations in Economic History 41, 1-25.
55. Roover, Raymond de, 1963. The Organization of Trade, in: M. M. Postan et al. (eds.), TheCambridge Economic History of Europe, Volume 3, 42-118.
56. Rose, Susanne, 2011. Wine trade in medieval Europe, 1000 -1500. London: Continuum.
57. Roth, Alvin E. and Marilda A. Oliveira Sotomayor, 1990. Two-Sided Matching, A Study inGame-Theoretic Modeling and Analysis, Cambridge: Cambridge University Press.
58. Roth, Alvin E. and Xing, 1994. Jumping the Gun: Imperfections and Institutions Related tothe Timing of Market Transactions American Economic Review 84 (4), 992-1044.
59. Roth, Alvin E., 2007. The Art of Designing Markets, Harvard Business Review, October2007, 118-126.
45
60. Roth, Alvin E., 2008. What Have We Learned from Market Design? Economic Journal 188,285-310.
61. Roth, Alvin E., 2016. Who Gets What and Why: The New Economics of Matchmaking andMarket Design. Boston: Mariner Books/Houghton Mifflin Harcourt.
62. Roth, Johann Ferdinand, 1802. Geschichte des Nurnbergischen Handels, 4 vol., Leipzig:Adam Friedrich Bohme.
63. Roth, Alvin E., and Marilda A. O. Sotomayor, 1990. Two-Sided Matching: A Study inGame-Theoretic Modeling and Analysis. Cambridge: Cambridge University Press.
64. Rothmann, Michael, 1998. Die Frankfurter Messen im Mittelalter, Franz Steiner Verlag:Stuttgart.
65. Rust, John, and George Hall 2003. Middlemen Versus Market Makers: A Theory of Compet-itive Exchange, Journal of Political Economy 111, 353-403.
66. Santarosa, Veronica Aoki, 2013. Pre-Banking Financial Intermediaton: Evidence from aBrokerage Law Reform in 18th Century Marseille, Working Paper, Michigan university.
67. Schubert, Hartmut, 1962. Unterkauf und Unterkaufer in Frankfurt am Main im Mittelalter,Diss., Frankfurt.
68. Spulber, Daniel, 1996. Market Microstructure and Intermediation, Journal of EconomicPerspectives 10, 135-152.
69. Stasavage, Daniel, 2014. Was Weber Right? The Role of Urban Autonomy in Europe’s Rise,American Political Science Review Volume 108 (2) , pp. 337-354.
70. Stein, Walter (ed.), 1893-1895. Akten zur Geschichte der Verfassung und Verwaltung derStadt Koln im 14. und 15. Jahrhundert, 2 vol., Bonn.
71. Verlinden, C., 1965. Markets and Fairs, in: Postan et al. (eds.), The Cambridge EconomicHistory of Europe Vol. 3, Cambridge: Cambridge University Press, 119-56.
72. Wee, Hermann van der, 1963. The Growth of the Antwerp Market and the European Econ-omy, 3 volumes, Leuven: Den Haag.
73. Wolf, Armin, 1969. Die Gesetze der Stadt Frankfurt am Main im Mittelalter, Frankfurt a.M.
74. Yavas, Abdullah, 1992. Marketmakers versus Matchmakers, Journal of Financial Intermedi-ation 2 (1), 33-58.
75. Yavas, Abdullah, 1994. Middlemen in Bilateral Search Markets, Journal of Labor Economics12 (3), 406-29.
46