Coagents as intermediaries in the book industry

This study investigates the use of coagents in the book industry. To reach international markets, domestic publishers typically license a title ’ s rights to third-party international publishers, a practice known as ’ selling rights ’ . Rights sellers can either choose to work directly in a local market or with intermediaries known as coagents. Using a data set of over 2000 international rights sales for Australian-authored titles of a major publisher, we examine whether employing the services of a coagent leads to superior outcomes as measured by the size of the advance. We find a positive relationship between the use of coagents and the size of the advance. This result is robust to a variety of contract lengths and international market conventions. The evidence suggests that, on average, coagents help secure better outcomes for authors and domestic publishers. More broadly, our study suggests that international intermediaries may play an important role in the export of cultural goods.


Introduction
To reach international markets, domestic book publishers typically license a title's rights to third-party international publishers, a practice known as 'selling rights'.This approach opens up new reader bases and revenue streams, contributing to an author's broader success beyond their home market.In smaller markets, rights sales serve as an important income source to supplement domestic earnings. 1Rights sellers can either choose to work directly in a local market or with intermediaries known as coagents.As we discuss in detail in the following section, coagents provide potential benefits via their knowledge of the local market and reduced transaction costs.In exchange for such service, they charge a commission of the advance they secure, along with a claim on potential royalties.
In this paper we examine whether coagents secure better advances using a data set covering more than 2000 international rights sales of Australian-authored books of a major multinational publisher.Our econometric results reveal a positive relationship between the involvement of a coagent and the size of the advance.We include a range of control variables related to the deal, country, book, author, and genre.Furthermore, we conduct various robustness checks to confirm our main result.Based on the implied probabilities from our preferred model, back-of-the-envelope estimates suggest the advance in an international rights sale with a coagent is, on average, 82% larger than without a coagent.
The remainder of the paper is structured as follows.In Section 2 we develop a theory of coagents as intermediaries, which provides a simple testable hypothesis concerning the relationship between the use of coagents and advances.The theoretical framework is also supported by a simple mathematical model provided in the Appendix.In Section 3 we describe features of the data set.In Section 4 we present our empirical findings based on the econometric results, with accompanying robustness checks.Finally, in Section 5 we summarise and discuss how the results inform theory related to intermediaries.We also reflect on shortcomings of our work and opportunities for future research.

A theory of coagents as intermediaries
Domestic publishers, literary agents, and authors typically sell physical books internationally by licensing rights to international publishers.2Rights sales grant an international publisher permission to edit, market, translate (if the title is being acquired for a non-English market), print, and sell copies of the original title.Rights sales require the domestic seller to pitch titles to acquiring editors in international markets.Successful pitches require a knowledge of local market trends and potential acquisition editors to approach.
Rights sellers can either work directly in a local market or with intermediaries known as coagents.Coagents work on behalf of rights sellers, with a focus on their local market (i.e. a French coagent is likely to be based in France).Coagents meet regularly with local publishers, gaining specialised knowledge of local demand.They are also able to pitch a domestic title in the relevant language.Rights sellers therefore typically have a large network of coagents to represent them in a variety of territories.
Coagents fulfil a comparable role to trade intermediaries, which are pivotal in international trade.They provide access to information and networks (Rauch and Watson, 2004), while also reducing asymmetric information and transaction costs (Biglaiser, 1993;Hendershott and Zhang, 2006).Additionally, intermediaries permit market access without the expense of establishing foreign subsidiaries (Felbermayr and Jung, 2011).Ahn et al. (2011) highlight the extra importance of intermediaries in challenging markets.
In markets with heterogeneous goods and idiosyncratic consumer preferences, intermediaries can facilitate connecting individuals with suitable products.In a theoretical setting, Johri and Leach (2002) show that the anticipated gains from trade between a consumer and a middleman surpass those between a consumer and a producer in such markets.Intermediaries have been shown to produce benefits in a variety of industry settings.In a study of real estate brokers, Bernheim and Meer (2013) observe that listings not associated with brokerage services lower the selling price of the average home by 5.9% to 7.7%.In the travel industry, Clerides et al. (2008) find tour operators accumulate specialised knowledge about hotels, which leads to more efficient market outcomes.
Intermediaries in cultural industries serve similar roles to those of other industries.Examining the film industry, Smits (2019) proposes that agents fulfil three vital functions: search and selection, coordination, and representation; with each function directly influencing cultural flows.Foster et al. (2011) note intermediaries in cultural markets utilise extensive social networks to manage the search and selection.Lizé (2016) observes cultural intermediaries largely originate from the middle class, exhibiting entrepreneurial tendencies, organisational prowess, and extensive practical expertise.
In the book industry, Caves (2003) notes that being represented by an agent, who acts as both a gatekeeper and an intermediary, can alleviate publishers' concerns regarding an author's work.Agents with a history of consistently poor author selection won't sustain their presence in the industry due to their recurrent interactions with publishers.On the other hand, owing to the impact of digitisation, some argue there is a reduced role for cultural intermediaries.For example, Peukert and Reimers (2022) highlight improved ability of advances to predict ex-post success in the book industry, owing largely to better data and consumer analytics.Such developments offer improved resource allocation and potential cost savings for publishers.However, despite these advancements, intermediaries maintain a pivotal role in the book industry, potentially due to entrenched cultural and legacy effects.Gehrig (1993) notes the presence of intermediaries hinges on prevalent beliefs and strategies.
A typical rights deal involves the payment of an advance with subsequent royalties paid after the advance has 'earned out'.Consequently, the size of the advance is a customary metric used to assess the success of rights sales.Publishers and literary agents receive a commission on all rights sales.While this amount may vary from deal to deal, industry insiders point to a standard commission of 20%.If the deal is secured with the use of a coagent, the coagent receives an equal share of the domestic publisher/agent's commission (i.e.10% for both parties) (Rudell and Rosini, 2017).Given diverse global markets, publishers (and authors) are likely to achieve better advances by employing coagents who generally have superior knowledge of local markets.This leads to the following simple hypothesis: Hypothesis 1.There is a positive relationship between use of a coagent and size of advance.
As noted above, beyond being able to mitigate information asymmetry by achieving better advances, coagents are also likely to provide other implicit benefits, such as reduced transaction costs related to finding the best deal.Of course, both the explicit (larger advance) and implicit (lower transaction costs) benefits come at the cost of shared commission.Because these benefits and costs differ between markets and across deals, it is not a-priori the case that coagents will always be used for a rights sale.
Intuitively, if the (expected) combined benefits exceed the cost of engaging a coagent, then it makes sense to do so (and vice versa if costs exceed benefits).We provide a simple theoretical model to illustrate potential equilibria in the Appendix (see Theorems 1-3).Furthermore, as noted by Corollary 1 in the Appendix, differences in expected net benefits can also explain why coagents may never, always, or sometimes be used at the country level.

Data summary
We use data from a major multinational publisher that includes all international rights deals for Australian-authored books made between 2008 and 2018.We observe a total of 2035 international rights deals across 57 countries and territories (countries, hereafter), covering 875 unique titles and 400 unique authors.In addition to deals covering multiple countries, we note a small number of deals were made multiple times for the same country in 148 (approximately 7%) of all observed deals.
Across the full set of 57 countries, deals in 12 countries never involved a coagent and deals in 40 countries always involved a coagent.Deals in the remaining five countries sometimes involved a coagent. 3As we discuss below, the distribution of different deal types over countries has implications for our empirical exercise.Table 1 reports the frequency distribution across the different ranges of advances for deals with and without coagents.As a weighted average using the midpoint of each advance range, 'no coagent deals' yield A$8580, whereas 'coagent deals' yield only A$4240.The distributions across the 19 genres we observe are provided in Fig. 1.In terms of genre frequencies, the most common genre was 'middle grade' accounting for 930 of the total observed deals (331 unique titles), followed by 'picture book' accounting for 345 observations (152 unique titles).
Table 2 provides summary statistics for the main continuous variables used in the regression models, where 'Advance' is simply the seven categories in Table 1 arranged ordinal 1-7.Although somewhat meaningless, the average for 'coagent deals' is again lower than 'no coagent deals'.The combined evidence may seem to suggest the latter type of deals outperform the former, but this ignores important confounding effects.For example, many of the 'no coagent deals' relate to longer contract terms, which are positively correlated with advances. 4As far as the 'deal gap' between the year of the rights deals compared to the domestic publication date, most deals take place within 10 years of publication (Fig. 2a), with the exception of some outliers.With respect to 'term of contract' most stipulate between 5 and 10 years, with the mode being five (Fig. 2b).However, some countries stipulate deals according to local terms of copyright, which are either 50 years (France) or 70 years (UK, US, North America).

Empirical analysis
In this section we investigate whether coagents are successful at achieving greater advances, in line with Hypothesis 1. Table 3 provides the main results of our empirical exercise using ordered probit models.In all models, the dependent variable is the ordered categorical variable relating to the different advance ranges as defined in Table 1.Our primary explanatory variable of interest, 'coagent deal' is a dummy variable taking the value one if a coagent was involved in securing the deal.As control variables, we include 'deal gap' as a continuous variable and 'repeat deal' as a dummy variable.We also include 'country', 'contract term', 'title', author', and 'genre' as sets of fixed effect dummy variables. 5he results from model (1) with only the inclusion of 'country' fixed effects and no other controls reveal a positive relationship between the use of a coagent and size of the advance.Model (2) includes additional controls for 'deal gap', 'repeat deal', and 'contract term' and the statistically significant positive association between coagent and advance size remains strong.In models (3), (4), and (5) we include 'title', 'author', and 'genre' fixed effects, respectively.These additional control variables address potential endogeneity that may extend from omitted variable bias if there is a systematic relationship between these variables and advance size that is also correlated with the coagent variable.We note the positive relationship between coagent and advance size remains strong, with estimated coefficients of similar magnitudes across all specifications.
While 'title' fixed effects provide an appealing remedy for potential endogeneity, we note a large number of singleton fixed effects (500 of the 875 titles in the sample), implying the average (median) observations per title is only 2.14 (1.00).A similar issue exists with authors given there are 400 authors, implying an average (median) of 2.18 (1.00) titles per author.Combined with a dependent variable with limited variation therefore provides challenges for estimation.Another problematic feature of the data is the lack of coagent variation for most of the countries in the sample.By using country fixed effects in the empirical models, this implies identification is largely driven by the heterogeneity in the five countries where we observe such variance. 6hile both 'title' and 'author' fixed effects provide useful empirical insights, our preferred specification employs only 'genre' fixed effects to mitigate omitted variable effects.While not as a-priori appealing as 'title' or 'author' fixed effects, there are econometric benefits from the coarser groupings due the reasons discussed above.Moreover, we know there is a skew towards certain genres in the sample that is helpful for estimation.In particular, 'middle grade' and 'picture book' represent 45% and 17% of all observed right deals, respectively.In any case, all models of Table 3 reveal a positive and statistically significant relationship between the use of a coagent and the size of the advance.
Using our preferred specification of model ( 5), we next calculate conditional predicted probabilities of each type of outcome across the different advance categories, which are reported in Table 4.With the exception of the 'less than $1000' category, coagents generate a higher probability of each outcome.Based on the implied probabilities and using the mid-point of each range, back-of-the-envelope Notes: All advance amounts in Australian dollars (AUD).  Notes: 'Advance' measures author/publisher advance (AUD) on a 7-point scale, where 1 = less than or equal $1000; 2 = $1001 -$5000; 3 = $5001 -$10,000; 4 = $10,001 -$25,000; 5 = $25,001 -$50,000; 6 = $50,001 -$100,000; 7 = $100,001 -$250,000; 'Contract term' is the length of the contract in years; 'Deal gap' is the number of years between the signing of the rights deal and the original year of publication.

P. Crosby and J. McKenzie
estimates suggest the expected advance with a coagent is A$6651 and without is A$3645, or an increase of 82%.
We interrogate the robustness of our main result in Table 5. Models (1) and (2) restrict attention to the 'contract term' groups of five and seven years, respectively.This stratification permitted inference abstracting from possible confounding effects related to contract length.Models (3) and (4) focus only on deals done with North America and Brazil, the two countries/territories with the most deals that sometimes used coagents.Model (5) incorporates all five countries that sometimes used coagents.These models permit inference in countries where there is no convention one way or the other in the use of coagents.All models support the main result and support Hypothesis 1.

Summary and concluding remarks
We find evidence of a positive relationship between the involvement of a coagent and the size of the advance for international rights sales.Our results are robust to deals of varying contract length, as well as different country conventions.Notably, in countries/territories that sometimes use coagents (including North America and Brazil, among others) we find convincing evidence of the positive relationship controlling for deal gap, repeat deals, contract term, country, and genre effects.Back-of-the-envelope estimates suggest deals with coagents attract (expected) advances approximately 80% greater than deals without coagents.
While coagents do, on average, increase the advance size, we note that this will vary between markets and deals.Additionally, we note the implicit transaction costs saved by employing a coagent will also vary between markets and deals.The total benefit will ultimately have to be compared to the cost of using a coagent (i.e.foregone commission) on a case by case basis.It is precisely for this reason that coagents are not universally used across all countries or even within specific countries.
We note several shortcomings of our empirical analysis that could be addressed in future work.First, the data are not ideally suited to inferring causal inference given the number of countries that either never or always use coagents.Such lack of variation restricts identification to a relatively small number of countries.Furthermore, while we include various controls, it is impossible to entirely rule out unobserved heterogeneity that may co-determine use of coagents and advances in specific countries.We therefore refrain from claiming to have identified a robust causal relationship between our main variables of interest.
Another limitation of our work is that our data are limited to just one multinational publisher, focusing only on Australian-authored books.Further research is necessary to determine if the findings of this study apply to other publishers, encompassing smaller imprints and various territories.We also note we only observe advances and not sales.While anecdotal evidence suggests many book deals never achieve sales that return royalties beyond earning out the initial advance, examining the relationship between use of coagents and sales is a natural potential extension to our work.Finally, and not least, we also only observe advance ranges and not specific amounts.The precision of all results and associated inference would obviously improve with precisely defined advance information.Notes: All models are ordered probits with the same variables as Table 3. Models (1) and ( 2) include only books subject to 5-and 7-year contract terms, respectively; models (3) and ( 4) include only deals made with North America and Brazil, respectively; model (5) includes deals made with countries where some deals are done with coagents and some deals are done without.Standard errors in parentheses; * p < 0.10, ** p < 0.05, ***p < 0.01.
P. Crosby and J. McKenzie

Fig. 2 .
Fig. 2. Deal gap and term of contract.Notes: Figure (a) is a histogram of number of years between international rights deal and year of original publication.Figure (b) is a histogram of terms of contract years, including deals done for country-specific 'terms of copyright'.

Table 1
Summary statistics.

Table 2
Summary statistics.

Table 3
Ordered probit results.Coagent' is a dummy variable, taking value one if the international rights sales involved a foreign coagent; 'Deal gap' is the number of years between the signing of the rights deal and the original year of publication.'Country FE', 'Contract term FE', 'Title FE', 'Author FE', and 'Genre FE' are fixed effects for country (or territory), length of the contract in years, title, author, and genre, respectively.Standard errors in parentheses; * p < 0.10, ** p < 0.05, ***p < 0.01.
Notes: All estimates based on model (5) from Table3.P.Crosby and J. McKenzie