Copyright Board Canada

Commission du droit d’auteur Canada

 

Date	2018-01-19
Citation	CB-CDA 2018-009
Regime	Collective Administration in Relation to Rights Under Sections 3, 15, 18 and 21 Copyright Act, subsection 70.15(1)
Members	The Honourable William J. Vancise Mr. Claude Majeau Mr. J. Nelson Landry
Proposed Tariffs Considered	Elementary and Secondary Schools – 2010-2015

Statement of Royalties to be collected by access copyright for the reprographic reproduction, in Canada, of works in its repertoire

Reasons for decision

I. INTRODUCTION

[1] On February 19, 2016, the Board issued its decision pertaining to the reproduction in Canada (excluding Quebec) of works in Access Copyright’s repertoire by elementary and secondary educational institutions and persons acting under their authority for the years 2010 to 2015. Access sought judicial review of that decision. On January 27, 2017, the Federal Court of Appeal agreed with Access that the Board failed to consider that expert evidence had been filed to estimate the degree of the underestimation of the number of copied works comprised in Access’ repertoire. The Court concluded that this was a reviewable error and sent the matter back to the Board for reconsideration. The present reasons represent the Board’s reconsideration of the matter, as mandated by the Federal Court of Appeal.

II. THE ISSUE OF REPERTOIRE

[2] Since Access can only claim royalties in relation to works that are in its repertoire and since Access and the Objectors did not agree on how this should be determined, the issue of which works were in Access’ repertoire was live during the proceedings.

[3] On June 3, 2013, the Objectors wrote to the Board, stating that

[i]n order to both adequately answer the claims of Access Copyright [with respect to repertoire,] and to make their own Case, the Objectors will need to conduct their own analysis on these data.

[…]

This repertoire analysis ‒ the matching of apparently published works to Access Copyright's claimed repertoire ‒ is vital for the Objectors’ own analysis in the current proceeding, but can only be carried out by Access Copyright. [underlined in original]

[…]

As the Board is aware, the Objectors do not have access to the database Access Copyright uses to match items in its repertoire. Such being the case, any repertoire analysis relating to the 2006 Survey, which the Objectors now require in order to prepare for the April, 2014 public hearing, must be undertaken by Access Copyright. The Objectors cannot do this work themselves. [1]

[4] On June 6, 2013, Access wrote to the Board, replying that

[t]he data generated from the [Volume Study] was provided to both sides for analysis in November 2006. In order to interpret the data, a ‘codebook’ was prepared, a further copy of which is attached to this letter as Appendix ‘B’. Using this codebook to interpret the data, in combination with the publicly available (on Access Copyright’s website) Access Copyright Exclusions List and the Access Copyright Repertoire Look-Up Tool, the Objectors have the means to carry out whatever independent analyses of the volume data that they may wish to carry out. To represent that “[this analysis] can only be carried out by Access Copyright” (see the third paragraph of the Objectors’ counsel’s June 3 letter) is simply incorrect. [2]

[5] The codebook provided the following:

Access Copyright Photocopy Volume Study Data Codebook produced on November 23, 2006

Cahier de codes pour les données de l’enquête de volume des reproductions d’Access Copyright produit le 23 novembre 2006

Note: this data set contains 384,227 records. The version used by Access Copyright splits some transactions related to the creation of overhead transparencies; therefore, it contains more records (albeit totalling to the same volume of copying) and it has some duplicate telkeys.

Note: Le jeu de données contient 384 227 observations. Dans la version utilisée par Access Copyright, certaines transactions ont été décomposées à des fins de création de documents pour rétroprojection. La version d’Access Copyright contient donc un nombre supérieur d’observations (qui représentent toutefois le même nombre total de reproductions) et certains codes de référence y apparaissent en double.

Access Copyright Photocopy Volume Study Data Codebook

Field name   Champ	Content   Description	Data type / format   Type / format de données	Discrete Values   Valeurs discrètes
Ac_pub_affiliate	Indicator that the publisher has a signed affiliation agreement with Access Copyright   Indique que l’éditeur a conclu une entente d’affiliation avec Access Copyright.	Binary         Binaire
Ac_rro_bilateral	Indicator that the publication is from a country with which Access Copyright has a bilateral agreement   Indique que la publication vient d’un pays avec lequel Access Copyright a conclu une entente bilatérale.	Binary           Binaire

[6] In other words, Access replied by stating that the Objectors could perform their own repertoire analysis, and could use as one of their sources of data for such an analysis the dataset, in conjunction with the codebook provided to them by Access. At this point, there was no suggestion by Access that there were significant coding errors, and that the data could not be relied upon.

[7] The hearing of the matter took place between April 29, 2014, and May 9, 2014, with closing arguments on September 12, 2014. Access did not raise the possibility that there were significant errors in the dataset at any point in the hearings.

[8] In order to better understand the evidence that was being discussed during the hearing, on June 6, 2014, the Board put certain technical questions to the parties. These questions included the following:

Please confirm that the correct interpretation of ac_pub_affiliate = 1 (as used in Exhibit Objectors-10 at para 101) is that the work’s publisher has signed an affiliate agreement with Access Copyright. Please confirm that the publisher is the owner of copyright of the work in every such instance.

[9] In its response [3] to the Board’s questions, Access stated that:

1. The code ac_pub_affiliate = 1 indicates that the publisher of the work has signed an affiliation agreement with Access Copyright.

2. It is apparent from a review of the 2006 Volume Study data (the “2006 Data”) that there are many other books, magazines and newspapers in the 2006 Data that have been coded as either ac_pub_affiliate = 0 or have a blank entry under ac_pub_affiliate for which the publisher, has in fact, signed an affiliate agreement with Access Copyright. As a result, using the “ac_pub_affiliate” field to measure the volume of copying of works owned by Access Copyright’s affiliates greatly underestimates the volume of such copying. A similar conclusion is apparent with the “ac_rro_bilateral” code, as explained in our answer to question 1(b) below.

[…] if the Board declines to apply the doctrine of issue estoppel, or finds that Access Copyright is not entitled to claim compensation for works by virtue of an agency relationship, it will be necessary to conduct a more detailed analysis of the data to delineate between books, magazines and newspapers that are claimed by virtue of affiliation and those that are claimed under the agency relationship. (emphasis added)

[10] This was the first point at which Access mentioned the existence of significant coding errors in the dataset. According to AC-112, these coding errors have the effect of underestimating the volume of copying of certain genres of works (books, newspapers, and magazines) whose owner of copyright was an affiliate of Access or whose owner was affiliated with a Reproduction Rights Organization (RRO). Despite this, in AC-112, there is no mention of the kind of review that occurred to identify the coding errors nor of any attempt to explain how such errors may have occurred.

[11] After the Objectors provided their own response to the Board’s questions of June 6, 2014, Access provided a reply thereto [4] in which it submitted a new report by its expert, Benoît Gauthier. This report provided a description and analysis of a new study performed by Access in preparation for the reply. The study considered “a random sample of book, magazine and newspaper transactions that had both ac_pub_affiliate and ac_rro_bilateral set to 0, that were claimed as part of its repertoire, not excluded and that were not made by Ministries of Education.” [5]

III. REASONS OF THE BOARD AND JUDICIAL REVIEW

[12] In its reasons of February 19, 2016, the Board wrote that “Access has provided no evidence of the degree of underestimation.” [6] On judicial review, the Federal Court of Appeal concluded that “the Board, through oversight, overlooked the expert evidence and submissions it accepted as exhibits AC-114 and AC-114A on December 5, 2014.” [7]

[13] The Court stated that it “should intervene and require the Board to assess the impact, if any, on the volume of compensable exposures.” [8] It consequently referred the matter “back to the Board for reconsideration of only the issue concerning the impact of the coding errors on Access Copyright’s repertoire.” [9]

IV. RECONSIDERATION PROCEEDINGS

[14] Pursuant to the Order of the Federal Court of Appeal, on August 30, 2017, in Ruling 2017-084, the Board ordered that the Objectors would have an opportunity to respond to AC-114. Furthermore, to ensure that such a response could be meaningful, the Board also ordered that

the Objectors may request data from Access related to the issue of coding errors, including data used in Access’ repertoire reanalysis. Access will take reasonable steps to provide such data. In the case of any dispute, either party may apply to the Board to resolve the dispute.

[15] On September 13 and 26, the Objectors requested “Access Copyright to provide any and all documents that prove Access Copyright’s claims that the Claimed Works were part of its repertoire throughout the years 2010-2015.” They argued that

[d]ue to the apparently large proportion of errors in Access Copyright’s repertoire analysis, the Objectors request the opportunity to double-check the repertoire status of works involved in both the reanalyzed transactions and in all other transactions that had originally been coded as involving affiliated works.

To do this, the Objectors need Access Copyright to provide them with repertoire information relating to all works in the sample that Access Copyright has coded as being affiliated, including those works that Access Copyright originally coded as being in its repertoire.

The Objectors refer to these, collectively, as “Claimed Works”, meaning all works that the Board found in its Decision of February 19, 2016 to be in-repertoire through either direct affiliation or affiliation through a foreign reproduction rights organization, plus all works that the Board originally found to be outside of Access Copyright’s repertoire but that Access Copyright claims in Exhibit AC-114 are in-repertoire.

[16] On September 27, Access replied, submitting that

the Objectors are attempting to enlarge the scope of the reconsideration of the evidence to be considered beyond what the Federal Court ordered (Circum’s report/AC-114A and its data/AC-114B). Any attempts to do so or attempts to re-litigate other evidence that was before the Board when it made its decision should not be entertained. The Court’s and the Board’s rulings are clear and settled.

[17] If Access is correct, and only the data related to the transactions considered in AC-114 may be considered, then we require no further submissions and can proceed directly to consider “the weight, if any, to be given to Circum’s report.” [10]

V. ANALYSIS

A. The Reanalysis Study

[18] The description of the study in AC-114A reveals a fundamental limitation: it can only identify errors in one direction. That is, by reevaluating only those transactions where the data indicated that the copying was from a work whose owner of copyright was neither a direct affiliate of Access (ac_pub_affiliate = 0), nor an affiliate of an RRO with which Access had a bilateral agreement (ac_rro_bilateral = 0), this analysis could only result in an increase to Access’ repertoire. If there were coding errors in the overall dataset that, when corrected, would decrease Access’ repertoire, these could not be detected by such a study.

[19] This makes such an analysis unreliable as an indicator of the net amount of underestimation. While the reanalysis may be a reliable indicator of the actual status of transactions initially coded as not being in repertoire, it is not a reliable indicator of the actual status of transactions initially coded as being in repertoire—and therefore not a reliable indicator of the net amount of underestimation. Any attempt to measure net underestimation would usually require sampling from the entire population of transactions, including those that were initially coded as being in repertoire.

B. Possible Contextual Information about Coding Errors

[20] While the reanalysis is flawed in its design, we nevertheless consider below whether there is sufficient evidence before us that could provide, through context, at least a non-quantitative sense of its reliability.

[21] There is insufficient contextual information that would indicate that there are reasons to believe that the reanalysis is a reliable indicator of net underestimation, such as information about the manner in which the coding errors were initially identified (e.g., were all transactions reviewed? Were only “double-0” transactions reviewed?).

[22] One such possibility was raised by Access when it argued that its reanalysis of books, newspapers, and magazines leads to a rate of copying of works not in repertoire (approximately 6.9 per cent) comparable to that of consumables (approximately 6.2 per cent); and that the similarity in these rates are an indication that the reanalysis better represents its repertoire than the initial data. However, setting aside that the means in which these groups being compared were constructed is significantly different, careful scrutiny of the data demonstrates that the only way to arrive at this result is to ignore the separate genres, and compare an amalgam of books, newspapers, and journals against a single genre, consumables. As Table 2.1 of AC-114A shows, the likelihood that a transaction previously identified as not being in repertoire remaining not in repertoire varied from 9 per cent for newspapers to 39 per cent for books, to 47 per cent for magazines. We therefore do not conclude that this indicates some kind of support for the proposition that the reanalysis is a reliable indicator of the status of all books, newspapers, and journals.

[23] Access provided some explanation for why significant errors may have occurred (Access did not believe it was “strictly necessary to precisely code and identify its affiliates in the Volume Study data.” [11] ), but it did not provide sufficient information that would indicate that the initial coding process would lead to certain kinds of errors more likely than others. For example, AC-114 claims that transactions representing approximately 28 per cent of the volume previously coded as not being in repertoire actually had insufficient information for affiliation status to be determined. Without sufficient information on how the initial error occurred, it appears very likely that similar errors would have occurred in cases where transactions were coded as being in repertoire.

[24] We note, for clarity, that in AC-114, Access noted the difficulty of establishing the identity of a work for a significant number of transactions, and that such transactions had no entry in the “ac_rro_bilateral” or “ac_pub_affiliate” fields. In its decision, the Board did not treat such transactions as not in Access’ repertoire, but interpreted them as “Unknown”, [12] and inserted them pro-rata into the other types. [13]

[25] There is also insufficient information that would permit us to conclude that the reanalysis should be expected to be more accurate than the initial analysis (e.g., was there a methodical error throughout the dataset that was located and removed? What was done differently, besides doing a “better” job?).

[26] The information sought by the Objectors pursuant to their request of September 13 and 26, would likely have provided a quantitative measure of the study’s reliability, and perhaps even provided some qualitative context to the identified coding errors. It is even possible that such information may have even demonstrated that there was a net overestimation of the volume of copying from Access’ repertoire. However, Access would not provide this information to the Objectors, instead arguing that the Objectors’ request is beyond the scope of the Order of the Federal Court of Appeal.

VI. CONCLUSION

[27] We are convinced that coding errors did occur. We note, for example, the eight transactions identified in AC-112 that were incorrectly coded as both ac_pub_affiliate = 1 and ac_rro_bilateral = 1, indicating that the owner of copyright of the work is both an affiliate of Access and an affiliate of a RRO, which is not possible under normal circumstances. This is normal, as such a large exercise as that performed by Access in 2005-2006 almost inevitably results in some errors. Without a reason to believe otherwise, one expects errors in a sufficiently large dataset not to be biased in one direction or another.

[28] Thus, on the evidence before us, it is not possible to conclude that the coding-error rate was different for transactions with a work initially coded as not being in repertoire than for those transactions with a work initially coded as being in repertoire. As such, we cannot give the reanalysis in AC-114 any meaningful weight as a reliable indicator for the net underestimation and cannot accept Access’ submission that we can rely on the initial data that indicate that a transaction was of a work in Access’ repertoire, but cannot rely on those data that indicate that the work was not.

[29] We therefore conclude our reconsideration as follows: there is no reliable evidence of the degree of net underestimation in this matter. Even when we look beyond the fundamental issue of the manner in which the reanalysis presented in AC-114 was performed, there is insufficient evidence, either from filings, evidence led during the hearing, or in responses to the Board’s related questions, that would permit us to conclude, in a qualitative sense, that the reanalysis is a reliable indicator of the net underestimation of repertoire.

[30] While the Board’s initial position was that it would require additional steps in its reconsideration, such as submissions from the Objectors, the fact that there is no additional information beyond AC-114 to consider has made this unnecessary. Since we do not give those portions of AC-114 dealing with the claimed underestimation of the volume of copying in Access’ repertoire any significant weight, it is not necessary for the Objectors to provide further submissions.

[31] As such, and despite the presence of errors in the dataset (which, as noted above, is not unusual in such a large exercise), the initial data remains the best available source from which to estimate the total volume of copying from works in Access’ repertoire. We therefore cannot reliably make any adjustments to the royalty rates the Board previously fixed on February 19, 2016. The Access Copyright Elementary and Secondary School Tariff, 2010-2015 is therefore unchanged.

Gilles McDougall

Secretary General