Internet Service Providers Report


Awareness and Current Practice:

Large and small ISPs diverged in their awareness of "the current status of ISP copyright liability in Canada" (see Table 2). 100% of large ISPs indicated they were aware of the current status of Canadian copyright law, and the majority were aware of copyright law as it pertains to ISPs in other jurisdictions (US, EU, and Australia.) On the other hand, only 46% of small ISPs indicated that they were aware of the current status of Canadian copyright law as it pertained to ISPs, and very few indicated awareness of other jurisdictions.

This difference is reflected in recourse to legal expertise, where all large ISPs either had in-house legal counsel dealing with copyright liability (87.5% of large ISPs) or had sought external counsel for copyright liability (12.5%); in stark contrast, only 2% of small ISPs had internal counsel and only 19% had ever sought external legal advice.

Only a minority of firms were aware of subscribing to any form of insurance that included liability for copyright infringement. Typically, these policies had deductibles in the range of $100,000 per incident and included copyright insurance as part of professional liability insurance.

Three-quarters of large ISPs are currently following some version of a voluntary notice and notice system, versus only 58% of small ISPs. Clearly, NN is not currently universally applied by Canadian ISPs. In addition, NN, as currently practiced, has evolved from the original CAIP/CCTA/CRIA agreement in two ways. First, all ISPs now accept copyright infringement notices via email as well as in writing. Second, some ISPs no longer send a written (or email) acknowledgement back to the rights holder originating the copyright notice.4

Number of Copyright Infringement Notices

The number of infringement notices received by each ISP for all available years is presented in Table 3. Unfortunately, only large ISPs had what were judged to be reliable records on the number of notices received either on a monthly or yearly basis for any period longer than six months.5 Three major trends are evident from this table:

  1. The number of notices is following a generally upward trend since ISPs began collecting these data (2001). In terms of number of notices received per ISP, the annual growth rate is 77.8% per year for the period 2001-2005.
  2. The number of notices received by large versus small varies significantly. On average, large ISPs receive 3.3 times as many notices per high-speed (Cable or DSL) subscriber than do small ISPs.6
  3. There is considerable variance in the number of notices received across ISPs. For large ISPs, the ratio of the max to the min is more than 13:1, indicating that on a per-subscriber basis, some ISPs get more than 13 times as many notices as other ISPs. The range is even larger for small ISPs, a number of which report receiving zero notices.7 Thus, there is considerable heterogeneity in the number of notices generated per subscriber, both within and across size categories of ISPs. The implication of this difference is that ISPs may bear very different burdens in terms of adhering to a Notice and Notice regime.

Analysis of Copyright Notices across Subscribers

The number of notices received by ISPs was analyzed in two ways. First, all ISPs reporting the number of subscribers and infringement notices currently being received (as of any month in Q1, 2005) were pooled and analyzed using cross-sectional techniques. The objective here is to uncover what are the current drivers of copyright infringement notices.

There are fifty-one (51) firms that report the number of each type of subscribers (dial-up, Cable, and DSL) and the number of infringement notices as of 2005. The relationship between each type of subscribers and the number of infringement notices received was analyzed. Details of this analysis are presented in Appendix 1 below. Overall, two main results emerge. First, high-speed subscribers (Cable and DSL) generate more copyright infringement notices than do dial-up subscribers. Second, large ISPs receive more notices per subscriber than do small ISPs.

Analysis of Copyright Notices over Time

The evolving nature of copyright infringement notices over time was analyzed using econometric methods for panel data; the details of this analysis are presented in Appendix 2 below. Here the objective is to understand how the drivers of infringement notices have been changing over time. As noted above, the sample of firms able to supply notice data on a monthly basis is confined to a subset of the large ISPs.

As is apparent in Table 3, the total number of infringement notices received by ISPs is increasing over time, at an average rate of 77.8% per year. Analysis reveals that the effects of time differ, however, across subscriber types. In particular, the number of copyright notices received per high-speed subscriber (Cable or DSL) has been increasing over time at a steady rate whereas the number of copyright notices received per dial-up subscriber has been falling over time.

Overall Analysis of Notices over Time and Subscribers

Putting the two preceding analyses together, the estimated number of copyright infringement notices generated per 1000 subscribers per month is indicated in Table 7.8 Three main effects are apparent from this analysis:

  1. The number of notices generated per subscriber is higher for Cable and DSL subscribers than for dial-up subscribers. As of the start of 2005, the number of notices per subscriber is now approximately equal for Cable and DSL subscribers.9
  2. The number of notices per subscriber is significantly higher for large ISPs than small ISPs. On average, large ISPs receive 2.7 times as many notices per high-speed subscriber as do small ISPs.10
  3. The number of notices generated per high-speed subscriber has been increasing since 2001 at the rate of 25.3% per year.11
  4. The number of notices generated per dial-up subscriber has been decreasing since 2001 and it appears that, by 2005, dial-up subscribers have ceased to be a significant driver of copyright complaints for large ISPs.12

Sources of Copyright Infringement Notices

For the most part, ISPs did not keep records of the senders of infringement notices. However, two large ISPs were able to provide a sample of the sources of all notices for two recent time periods (2003-2005) spanning several months. These notices were grouped into categories, and the relative frequencies of these categories is listed in Table 8.

As can be seen, the vast majority of reported notices (96%) originated in the US, either from studios, software publishers, the Motion Picture Association of America (MPAA) or US firms acting on behalf of rights holders that were not revealed.13 Less than 2% of notices could be linked to directly to Canadian rights holders, although it is not known what proportion of notices is ultimately on behalf of Canadian rights holders. Unfortunately, the ISPs could not separate out notices by type of media concerned (e.g., music, movies, television). Since most Hollywood studios represent at least two if not all of these categories of media, it is impossible to classify the type of media within the notices received from these studios. However, qualitative assessment of respondents within both large and small ISPs formed a strong consensus that the vast majority of notices from US studios now relate to movies rather than to music.

Current Notice and Notice Practices

Table 9 describes the typical process in ISPs currently following the Notice and Notice regime. The left column describes the eight steps typically undertaken in a large ISP when a copyright infringement notice is received in reference to one of their subscribers. The right column presents the corresponding process in small ISPs, which typically comprises six steps. Overall, the process in the large ISPs reflected some investment in information systems and administrative procedures to support and/or partially automate the notice and notice practice.

In no firm interviewed was all of the information required to comply with notice and notice available from a single completely integrated source. Nearly all ISPs have to bridge two completely separate sets of information systems in order to respond to notices. These two types of systems are: (i) provisioning systems (those systems aimed at actually providing Internet access to subscribers), and (ii) customer billing systems. This challenge for ISPs stems from the fact that rights holders cannot identify individual infringers on the Internet, and must instead report an Internet Protocol (IP) address and date/time to the ISP responsible for that IP address. To the extent that it is possible to link an IP address assigned at a particular date and time to a particular subscriber account, a number of data sources must be searched and manually linked. The most crucial information in this process — logs of which IP address was assigned to which device at what time — is most frequently produced as a by-product of operational systems and stored in a "flat file" format. As a result, these files are not indexed and neither do they have any integrity constraints built into them. Searching these logs is thus time consuming and not guaranteed to be 100% accurate. The data sources used by ISPs are listed in Table 10.

For large ISPs, this process required, on average, 12.3 minutes of labour versus 47.4 minutes in small ISPs. This difference may reflect the fact that a number of large ISPs have engaged in some measure of systems development or integration to automate or expedite at least portions of the process. ISPs also reported the hourly wage for the personnel processed infringement notices, inclusive of benefits and taxes available. For large ISPs, the average hourly labour expense was $42.30, or 71 cents per minute; the comparable figures for small ISPs were $41.47 per hour or 69 cents per minute. Based on the average labour cost per minute and the average number of minutes per notice, the average cost of the Notice and Notice practice (where confirmation of notice need not be seen to the rights holder and neither is there a billing process in place) is estimated at $11.73 for large ISPs and $32.73 for small ISPs.14 The large difference in this number reflects the fact that many of the very small ISPs process a very low volume of infringement notices (perhaps 2–3 per year) and, on average, require a great deal of time to do so.

Another way to calculate the average cost of processing an infringement notice is to weight the cost for the ISP by the number of notices processed to produce an industry average per-notice cost. When this weighting is applied to large firms, it has little effect: $11.76 versus $11.73.15 That is, the cost of processing each notice within large ISPs is more-or-less fixed and doesn't vary with the number of notices processed.16 The weighted approach reveals a very significant difference for small ISPs, with the average cost dropping from $32.73 to $5.20. The conclusion to be drawn from this weighted comparison for the cost of notice processing at small ISPs is that firms that process more notices do so, on average, at a much lower cost per notice. That is, there appears to be an economy of scale in notice processing: the more notices processed, the lower the cost of each notice.

The fixed costs associated with notice processing were very difficult for respondents to quantify. These fixed costs include all of the investments undertaken in order for an ISP to conform to the Notice and Notice standard. These investments include the development of business process and administrative systems to perform notice and notice, development of databases or record-keeping systems to track notices, and systems development or systems integration work to automate or support the notice and notice process. The majority of large ISPs, as well as "larger" small ISPs, had undertaken either systems development of systems integration work of this sort. For example, a few ISPs have developed systems that automate or support the process of linking an IP address at a given date and time to a subscriber account or to link a subscriber account to a billing address. However, the vast majority of respondents were unable to quantify these expenditures; indeed, the number reporting is so small that confidentiality concerns prohibit reporting anything other than the approximate range of these figures, which was $50,000 to $300,000. Large ISPs, on average, assessed an "overhead" cost of $3.00 per notice, reflecting their estimate of these fixed costs spread out over the number of notices processed, as well as other overhead not included in labour costs (e.g., buildings and maintenance). Because small ISPs did not account for or estimate any overhead or fixed costs associated with notice processing, their estimates of the average cost of notice processing are likely underestimates. However, the magnitude of the underestimate is likely to be quite small because small ISPs typically had not engaged in systems development or integration, and neither did they have employees dedicated to notice processing.

Note again that there was considerable variation in the reported cost per notice. Confidentiality concerns make reporting the range of costs impossible, but the ratio between the highest and lowest costs within Large ISPs was nearly a factor of 10:1, and the range for small ISPs was even larger. For the most part, these differences in cost are idiosyncratic reflections of the choices made in the past — some many years ago — about the specific IT architecture and infrastructure implemented by each ISP. Since information systems work in a complex network of complementary components (e.g., computing hardware, operating system software, database systems, transaction processing systems, financial systems, middleware, networking hardware, etc.), the information architecture in a particular firm displays both lock-in due to switching costs and path dependency.17 Not only do the current costs of adhering to Notice and Notice vary considerably across ISPs, but the costs of integrating or automating these systems in the future will also vary considerably. Again, these variations in current and future costs imply considerable difference in the burden of Notice and Notice across ISPs.

Pricing and Implications for Consumers

The pricing practices of ISPs and their perception of the degree of price competition within the market is presented in Table 11. Although nearly all ISPs have changed their prices in the last three years, nearly all providers of Internet access priced their offerings very close to their competitors. There was a perception that perhaps only two firms in the industry were "price setters," with the remainder of firms being price followers. Nearly all ISPs offered some form of promotional pricing, and the perception is that the churn rate (subscribers switching ISPs) is very high in the market. One approach that large ISPs have taken in an attempt to reduce churn is to offer discounted pricing on "bundles" of service, typically consisting of some combination of Internet access, telephone service, Cable TV service, and wireless telephone service. While 100% of large ISPs had chosen to offer bundles, very few of the small ISPs were able to because almost none of these ISPs provided more than one of these services.

Given the competitiveness of the Internet access market, it appears likely that prices haven't increased as a result of notice and notice to this point in time. It is possible, however, that absent these notice and notice costs, prices of Internet access would have fallen more than they actually have.

In many large ISPs, the resources and people who handle infringement notices typically have other duties; in many organization the "abuse desk" handles copyright notices, as well as issues like SPAM, denial of service, inappropriate content, etc. Some ISPs suggested that the ability of their abuse departments to attend to these other issues may have been compromised as a result of the increasing number of copyright notices. Consequently, the quality of service experienced by subscribers may be lower than would have been otherwise.

4 Unfortunately, data on the number of ISPs sending acknowledgement back to the rights holder were not captured, so it is impossible to quantify this practice.

5 Data from small ISPs is much less accurate, since very few small ISPs actually maintain records of the number of notices received over time; instead, these numbers are largely based on the memory of the respondent.

6 The number of notices per high-speed (Cable or DSL) subscriber is calculated in Table 4. On average, large ISPs receive 7.7 copyright infringement notices per month per thousand high-speed subscribers whereas small ISPs report receiving only 2.3 notices per month per thousand high-speed subscribers.

7 Given the small size of some of these firms (fewer than 200 subscribers), this number is plausible.

8 These estimates are calculated using the sample means of infringement notices and subscribers. Given the presence of firm type effects in the cross-sectional model and time-effects in the panel data model, the preferred specifications in Table 7 are cross-sectional Model 3 and panel Model 4. The results are presentedin terms of the marginal effect of an additional thousand subscribers of each type on the number of copyright infringement notices received each month.

9 This comparison of the difference of Cable versus DSL subscribers is taken from panel Model 4, evaluated at January 2005.

10 For this calculation, the average values of high-speed subscribers are used for both small and large ISPs and the effect sizes are taken from Model 3 of the Cross-Sectional results. They are calculated as (3.45 + 4.62) / (0.73 + 2.28) = 2.68.

11 For this calculation, the average values of the interaction terms between month and Cable and DSL subscribers was taken from Model 4 of Table 6 and converted to an annualized value.

12 Although the estimated effect of additional dial-up subscribers as of January 2005 in the panel model is negative and significant, the most sensible interpretation is that dial up subscribers have no effect on the number of copyright notices received. Dial up subscribers remain a significant driver of infringement notices for small ISPs.

13 The common term for such firms, which seek out infringers and deliver notice to them through an ISP, is "bounty hunter."

14 Calculations based on figures from Table 9:
Large ISPs: Average Cost = 12.29 minutes * $0.71 per minute + $3.00 overhead = $11.73
Small ISPs: Average Cost = 47.43 minutes * $0.69 per minute + $0.00 overhead = $32.73 

15 Calculations based on figures from Table 9:
Large ISPs: Average Cost = 12.16 minutes * $0.72 per minute + $3.00 overhead = $11.76
Small ISPs: Average Cost = 16.77 minutes * $0.31 per minute + $0.00 overhead = $5.20

16 The total cost of processing notices, of course, depends directly on the number of notices: total cost is the product of average cost ($11.76 for large ISPs) and the number of notices processed (4426 per month for large ISPs as of Jan 2005) for a total average cost of approximately $38,900 dollars per large ISP per month as of January 2005.

17 See, for example, Shapiro and Varian (1999) for a discussion of the economics of information technology and systems.