Competition, innovation and intellectual property rights in software markets
Paul de Bijl, Michiel Bijlsma, Viktoria Kocsis - CPB Netherlands Bureau for Economic Policy Analysis
Public policy in relation to open source software: is public policy supporting open source desirable? It depends on the type of software and the market characteristics.
Traditional view: innovation is costly, and ex ante investments need to be recovered ex post through IP rights. Radically different model: innovation is interactive; IP rights should protect against knock-offs but allow building upon.
Proprietary model offers the best prospects for recovering investments, which may be important when you need to purchase expensive outside assistance—e.g., games, where you need to purchase graphic artists’ services. Follow-on innovation takes place through contracts/licensing.
Open source model is more interactive/organic. Need a different business model—support, education, complementary products alongside giving away product. Other incentives: job market signalling, intrinsic motivations. No barriers to follow-on innovation except that the follow-on has to use the same type of license structure.
Question: what kind of market failures are important to deciding on policy intervention? (1) Lock-in for consumers/vendors—consumer switching costs or network effects. (2) Which model is best for innovation. If development requires expensive expertise that has to be purchased outside the programming community, strong proprietary/closed licensing models may be better. Are there any obstacles to efficient contracting for follow-on innovation. If the first mover can efficiently contract, then there’s no compelling reason to stimulate open source.
What can the government do if open source is desired? Be lead customer for open source. Give preference to open source in procurement. If open source is not superior, government may still want to do increased competition measures more generally—focus on interoperability. Keep incentives in mind, but if there is serious customer lock-in, competition is a problem and can relatively risklessly be increased without harming incentives.
Government failure should be taken into account: picking winners is a risk. And bias in public procurement may distort incentives. Rent-seeking problems. Alternative solution for lock-in: enforce seamless exchange of files between users in network effect markets—text editors/office applications. Can enforce open standards.
Kathy Strandburg: What is the proper concern of governments? Could want freedom from being tied up with a big US company; could want specialized services/programs—it’s not just about encouraging open source for efficiency reasons.
A: Reason for this report: parliamentary debate over the cost of software for the government, argument about cheapness of open source/quality.
Copyright’s Latest Communications Policy: Content-Lock-Out and Compulsory Licensing for Internet Television
Marvin Ammori, Kate Aishton, Chris Riley
A media policy problem: the impediments to the mainstream emergence of online video. Goal: the internet should be a competitor to the plug that comes into your house for TV. Three media policy focused problems: blocking/discrimination against content; metering; set top boxes; and the big one, content lockout, the inability of internet providers to get access to the televised content that will drive adoption.
Blocking/content discrimination: net neutrality. Metering: video takes up lots of bandwidth—streamed Netflix movies will exceed the standard internet usage. Metering technique can impede adoption. Set-top boxes: provided by cable provider, and don’t integrate the internet so you can watch on your TV, which is the holy grail of integration because the TV is so much a bigger screen.
Big deal: content lockout. Even if you have the tech, there’s difficulty accessing the mainstream broadcast content, which will be required for effective competition. We want to see someone other than the cable companies and other than the TV companies to offer you that.
Specific problems with getting access to content: The Daily Show was available online, fullscreen, whole show. Time Warner didn’t like the competition and put together a list of all Viacom shows available online, and threatened to teach customers how to use these in order to drive down the program access fees/affect ad rates. While the terms they negotiated aren’t public, inside sources say if you look at the Comedy Central website now, you see that you don’t get the Daily Show intro, a tiny screen, and multiple clicks are required to watch.
Joost, a video startup, gave up serving consumers directly. Online content has a lot of movie trailers, some porn, other things that aren’t the must-have content needed to attract viewers. Even good content online is barred from displaying on your TV screen in an attractive way—fight between Hulu and Boxee is an example. Content producers own Hulu, which has a lot of great content, but Boxee’s software allowed consumers to stream online video to TV simply, with great format, and using your TV remote control. Last year, Hulu suddenly severed the relationship. They don’t want the online experience to mimic the cable experience.
As best as we can tell, it wasn’t fear of competition with Hulu, but fear of competition with cable TV, where broadcasters make a lot of money through licensing. Cable companies are using their market power and relations with content companies to discourage software companies from getting their hands on their valuable content. We are not saying we think this is right now illegal, but we think it is bad policy.
This comes out of the retransmission consent procedures—cable companies aren’t legally allowed to discriminate against competitors in favor of their own content. There’s also a meaningful nonexclusivity provision, DC Circuit just heard oral argument on this—can’t have an exclusive cable channel; though satellite companies are allowed to have exclusive channels. We’re starting to see massive market power among cable companies.
Solution: compulsory licensing. Allows competition and compensates copyright holders. Not a crazy change, but consistent with history of new technologies. Internet radio is a model that could work for online video. Sets out who has access to compulsory license, Copyright Royalty Arbitration Board for reasonable royalty if negotiations fail.
Q: How much reluctance is due to cable industry not noticing revenue online?
A: No substantial shift yet—we live in more urban areas with faster broadband, and people of our generation/class position tend to watch more stuff that’s available online: Daily Show v. Oprah. Also, the current model online does not produce the same amount of revenue as the cable market does. You can see where their concerns come from!
Q: TVonline—effort by cable providers to allow internet access to cable programming.
A: Better than nothing, but you need a cable subscription—impedes market entry by people who have no connection to the current cable industry.
Q: Why compulsory licensing instead of program access? ESPN 360.
A: He doesn’t think program access is done right. Compulsory licensing doesn’t eliminate troublesome legal structures, but its beauty is that it creates a more open market for people who want access to the content. Cable TV can benefit from compulsory licensing—smaller companies complain they’re screwed by big broadcasters.
Q: Internet radio licensing rates have been controversial—set rates so high that smaller players are driven out. How to handle that?
A: No perfect solution. If compulsory licensing is seriously pursued, we can’t repeat past failures, but framework is right.
Q: Is metering a nefarious conspiracy? Small percentage of users uses huge percentage of bandwidth.
A: It depends on what the numbers are. Comcast’s 250 gig soft cap—that’s probably going to allow a lot of HDTV. But TWC is using 50 gigs, and if you use the average household numbers for TV watching, that would triple/quadruple your monthly internet bill to watch the same amount of TV online. Plus this type of metering doesn’t decrease congestion at prime time.
Q: This is a big challenge to copyright rights. Is there something special about the internet and broadcast, when copyright owners can decide when to distribute movies on HBO or in theaters?
A: This isn’t a question of user rights to access content—it’s a matter of media policy, assuming copyright rights are in place. A lot of engineers, TV people think that 40 years from now all TV will be via IP. We should manage that transition properly, as a matter of media policy.
Characterizing Digital Media Exchanges in a University Campus Network
Alexandre M. Mateus, Jon Peha - Carnegie Mellon University
We don’t know as much as we should about p2p usage; users may refrain from disclosing their behavior, and it’s hard to tell how (if at all) filesharing affects copyright owners’ sales. Various tech has been considered, including deep packet inspection (DPI): deployed in some universities in the US. Considered by lawmakers as possibly mandatory for universities/ISPs. His study: Analyzed results of extensive DPI monitoring in a campus network.
Objectives: assess effectiveness of DPI, consider whether intervention is required, shed light on impact of p2p on revenues of copyright owners.
Over about 1 month, close to 40% of students used p2p and 70% of those transferred or attempted to transfer copyrighted content—4 titles per student per day. All demographics participated in filesharing, regardless of ages, genders, majors—even statistically significant differences were often practically small. P2p activity decreased about 10% over Spring 2007 to Spring 2008 (and 20% decrease in those using p2p detected as transferring copyrighted content), across demographics.
How to interpret these results? First, these are lower bounds because the tech was limited. It didn’t detect encrypted p2p traffic. It also didn’t detect activity involving copyrighted content if it wasn’t in the database used by the tech. Also, content transferred inside a zip file couldn’t be detected. And if the tech didn’t collect consecutive seconds of the content—video required a lot more bytes of collection before it could be ID’d. There could be a decrease in actual p2p use, or an increase in the use of evasion techniques.
In spite of these limitations, given enough monitoring time, DPI could detect most users using p2p to transfer copyrighted content (because people tended to transfer multiple files). You can tolerate a lot of false negatives and still detect most users. We can go as low as detecting 10% of events and still see about 90% of the users—doesn’t scale as well when you look at unique titles though.
Gathered all titles/filenames in metadata, regardless of whether it was in central database. Mostly songs, movies, and TV shows. Filenames: about 20% isn’t present in central database—adult content, software, books, pictures. Discrepancy between detection via metadata and via database, which is really good at songs. Huge discrepancy in video: movies/TV shows. 25% of users are transferring video if you look at file names, but only 2-4% are if you look at the database. So, can the database still find people who are transferring video, given the habit of transferring multiple files? For songs, 80% of all users transferring a particular popular song would be detected by the database, while 45% of all users transferring a particular popular video would be.
Due to predominance of video content, Bittorrent users are less likely to be detected transferring copyrighted content as compared to Gnutella.
Impact of unauthorized p2p transfers on content sales: compare p2p usage to iTunes usage, compared by IP address, which doesn’t represent users but is close enough because all IP addresses were fixed in the dorm rooms over this period. Broke down iTunes activity by number of bytes transferred. 1/3 of p2p users still use the iTunes store, and are somewhat more likely to go there only to sample content, but ¼ still buy from iTunes. Comparable % of p2p and non p2p users purchase content from iTunes. P2p and non p2p users download comparable quantities from the store as well. Contradicts the market harm hypothesis.
The figures on use, while decreasing, are compatible with RIAA estimates.
Limitations: encryption! DPI can only ID content in a central, constantly updated database, and only when it’s in the clear. Detection of video will be harder than detection of audio. But still, given weeks, you can still ID the users who are doing it.
Q from Strandberg: Bittorrent numbers are very low—10% of use detected as copyrighted content. Even accepting that video detection is limited, doesn’t that show high non-infringing use of Bittorrent?
A: He can’t be sure. In a single day, the numbers are lower, but they might get higher for an individual user over time.
Q: Bittorrent source files tend to be more compressed, .rar and other larger files.
A: Did you see increases in encrypted traffic? You could see source/destination—could you see whether it was coming from a tracker. Were you able to match video and see whether legal online availability decreased downloads?
Q: We have a slightly increased percentage of unidentified/unclassified traffic, but he can’t say that’s because of encrypted p2p. Source/destination: can’t do much given that they’ve anonymized the data. Also hasn’t done analysis on availability—but this was all collected prior to Hulu’s full launch, and he thinks that might have made a big difference. (Also prior to DRM-free music on iTunes.)
A: What was the school’s policy on p2p?
Q: Said don’t use it for illegal purposes; didn’t have bandwidth limitations; publicized that DPI would be occurring.