Monday, November 16, 2015

State Governors on Refugees are Like the FDA on New Drugs

The incentive structure that the FDA faces in approving new drugs is well known to be skewed in such a way that slows down innovation implementation of potentially life saving treatments. It goes something like this: If the FDA does not approve a drug that would have saved thousands of lives, no one will know. If the FDA approves a drug that kills a few people, everyone will know. Therefore, the incentives are such that decisions will skew towards caution even if a different policy would have saved more lives. The combination of availability bias and costless risk aversion yields a system which is detrimental to those who most need help, and it is this same combination that now threatens Syrian refugees.

Today seems to be official “governors against refugees” day. In the past several hours the governors of the majority of U.S. states have specifically come out in opposition to or banned by executive order their state taking in more refugees from the war in Syria citing security concerns and an inability to background check new entrants.

There are several problems with this logic. First, it is not that hard for a state to run background checks on a few thousand people. Connecticut Governor Dan Malloy (one of only two governors to support refugee acceptance today, the other being Peter Shumlin of Vermont) says that his state plans to continue to accept around 1,600 Syrian refugees and can easily complete background checks before entry. Second, states don't have have the authority to stop resettlement of refugees and there is not really any way to prevent refugees from moving between states even if states could bar them initially. Third, and most interestingly, the extreme precautionary principle being applied in these cases ignores economics and is detrimental to both natives and refugees.

The economic argument for open immigration is iron clad; it simply is the case that immigrants improve and do not harm the domestic economy in the country to which they move. Given this fact, there should be a pretty high burden of proof on those who want to impose restrictions on visas to refugees as doing so harms not only those fleeing the monstrous conflict in the Middle East but also the domestic economy.

Surely these governors think they have a pretty good reason in the form of security concerns. They don’t want an event like the Paris attacks to happen their state. I do not doubt the purity of their motives, but the nature of incentives and unseen costs makes the seemingly obvious case for caution more complicated. In short, overly precautionary restrictions on immigration have opportunity costs. But because these costs are not as visible as the potential security threat from allowing in a few bad actors, there is incentive to place an irrationally high value on safety.

No one will miss the additional innovation and production that could be brought about by allowing in refugees who will buy goods, pay taxes, and maybe even settle down and start businesses. All the economic benefits of immigration apply to refugees as well, but it is difficult to mentally account for what might have happened.

Terrorism, on the other hand, is obvious to the eye. The images and stories from Paris in the past few days provide firm footing for a call to greater caution. If we admit Syrians to the U.S. and one act of terrorism occurs, everyone will know. If we do not admit Syrians to the U.S. when doing so would have saved many lives and economically benefited both parties, no one will know.

This is not just an academic triviality; there are real lives at stake. In the case of the FDA, policy makers are beginning to understand the perverse incentives involved, and the adoption of “right to try” laws as increased. Unfortunately the same is not true of the refugee situation. Republicans have often been the ones in favor of reducing FDA control over patients’ decisions, but they seem to forgo this logic in the refugee issue; today’s loudest voices against refugee admittance, including many of the governors listed above and Presidential candidates Carson, Cruz, and Paul, have been from the GOP. The events in Paris seem to have shocked the rationality out of decision making, and the fate of many refugees will now be decided by the fears of policymakers rather than by measured composure that could save lives.

Availability bias is always a danger in making balanced security policy, and, in this case, state executives are allowing it to mercilessly drag them around. The opportunity costs of keeping out immigrants of any kind are very real and very harmful, but in the case of Syrian refugees they are even higher. Keeping these refugees out condemns them to horrible situations, and even if you don’t want to help these people, we should at least not slam doors in their faces.

Friday, June 19, 2015

Backward Reasoning on IP Protectionism

The folks over at IP Watchdog recently responded to a question from a reader who asked for proof that innovation would not happen in the absence of “sufficient IP protections.” As is often the case with IP advocates, the Watchdog writer thinks the case is glaringly obvious but chose to respond anyway because it is a “message worth repeating.”

Apparently she also thinks the case for intellectual property regulation is so obvious that it is not worth giving sound evidence. The arguments put forward are amazingly unperceptive and do not hold up to critical analysis.

The article begins with the standard argument of “economic fundamentals;” high R&D costs plus low marginal cost of replication and production means that innovation will not happen in the absence of patents. The jump is then quickly made to a pharmaceutical example to show the importance of patents in general. This is a misleading example because the question of IP protection is always an empirical one. Are fixed costs of development actually high enough and are marginal costs of duplication actually low enough that there will be a net loss of innovation in the absence of IP protection? The argument of myself and others is that they are not.

Pharmaceuticals are one of the only conceivable realm in which there is a possibility of the balance being so out of whack that some kind of IP protection scheme may be desirable. (Parenthetically, this does not mean that the present system of US patents is the most desirable form of protection. It is even likely that some non-governmental solution is preferable to any centralized regulation.)

The particulars of pharmaceutical patents aside, their uniqueness makes them poor candidates for analogies to the rest of the IP system. There is conceivably a case for IP protection of new drugs, but that has almost no relation to whether a farmer should be allowed to modify his tractor, and it is quite different from saying that patents on off the shelf Wi-Fi equipment should be able to be owned by patent assertion entities and asserted against small end users. Trying to set up pharmaceuticals as a blanket justification for all patents and copyrights as currently conceived is misleading at best.

The main thrust of the article is to argue that a correlation between countries with strong IP regulations and those with a lot of innovation. Even ignoring the dubious proposition of calculating an innovation score for individual countries, this reasoning is backwards. Innovation does not result from IP laws; IP laws result from rent seeking after innovation occurs. Every businessman would like to have a monopoly on the good or service he produces, and this desire, while usually frowned upon, finds social acceptance when couched in terms “intellectual property.”

But, of course, there must be inventions before inventors (or other IP holders) can lobby for monopolies. It should not be surprising that innovative countries have strong IP regulations because where there is much innovation there is much incentive to monopolize one’s piece of it. The presence of strong IP laws in innovative countries does not indicate that the IP laws where the cause of innovation any more than the presence of sick people at a hospital indicates that the hospital is the cause of the sickness.

Wednesday, May 20, 2015

Two Charts on Film Piracy

What follows is an excerpt of the results of a longer paper on The Effect of Increased Availability of Online Piratable Copies of Films on Box Office Revenue. The availability of screener copies online almost immediately for films released late in the year provides an exception to the rule of normal windowing, and the paper attempts to exploit this anomaly to discover whether the immediate availability of high quality piratable copies of films still in theaters affects box office revenue.

The films were first broken into two groups by year of release: Group 1 (1990-2002) and Group 2 (2003-2015. The groups were chosen to delineate between periods of growth in Internet film piracy, specifically the development of the protocol BitTorrent in 2003 (Dahaner and Waldfogel 2012). Before BitTorrent, film piracy via the Internet was more difficult because the large file sizes made it inconveniently slow. The new protocol, however, allows for much faster transfer of large files making downloading movies more feasible. These groups were then subdivided by month and the mean worldwide box office revenue number for each month was calculated. The result of these calculations is seen in Chart 1.

Recall that the relationship sought is the difference in revenue for end of year movies as piracy became more prevalent. The data do indeed show a decrease in revenue from November to December, but a drop occurred during between those months in both time periods. The magnitude of the change is especially relevant to this analysis; the change for each time period is shown in the table below.

Tuesday, May 19, 2015

How the Internet Changes Sharing

What follows is an excerpt of the results of a longer paper on The Effect of Increased Availability of Online Piratable Copies of Films on Box Office Revenue. The paper goes on to use the framework described below to conduct some analysis of the film piracy. The results are available in the the full paper and will also be part of a later post.

It is impossible to tell merely from the presence of sharing whether each participant values the good in question more than its full price; we can only say that he values the good more than the price required of him by the group plus the transaction costs associated with forming and participating in the group.

Transaction costs are an essential factor in examining whether and which groups form. These account for costs associated with gathering information, finding people who want to join a group, and determining what good or goods the group will consume. The magnitude of transaction costs can be diminished by technology, such as the Internet. If there is insufficient technology to overcome transaction costs, then sharing groups cannot form. For example, groups are limited by geography; I cannot easily share a DVD with someone 100 miles away from me without spending large amounts of time and money to meet up with him. Those transaction costs are high enough to keep sharing from occurring. That is, they are too high until transportation technology enables travel to take place at an affordable price, thus overcoming the time and money cost of traveling great distances.

Given all this, it is safe to say that sharing will occur when multiple people (with sufficiently low transaction costs and the technology to overcome them) each value a good less than its price, but the sum of how much they value the good exceeds its price. 

Sharing may also occur even when a participant values the good more than its price. Because each buyer wishes to minimize his opportunity cost, he will enter into a sharing-group if doing so will decrease his cost more than it would decrease the marginal utility of the good. This decrease in marginal utility will necessarily occur, when the good in question is rivalrous, because of transaction costs.

Given that people often act in groups to consume a good, there are two basic ways those groups can acquire that good. Either all (or some fraction) the members of the group pay part of the costs or one member pays the cost for everyone else.

All Members Pay (AMP) Groups

Before the proliferation of the Internet, transaction costs typically limited the size of sharing groups to a few people who already knew each other; e.g. you and your friend combine to buy a DVD. The cost of the DVD and the revenue to the seller is half of what it would be if you had each bought your own DVD. For any sharing group, the cost to the participants and the revenue to the producer is equal to 1/n  (where n the number of participants in the group) the cost and revenue if each member bought a separate instance of the good. While the producer would prefer greater revenue by selling a separate good to each member of the group, his net revenue is only decreased if more than one member of the group already valued the good more than its full price, meaning that they would have purchased it anyway if the group had not allowed them to minimize their cost.

One Member Pays (OMP) Groups

In some cases, groups involve one individual paying the full price for a good and sharing it with multiple other people who pay no portion of the price of the good. For example, one might buy a DVD and screen it for three friends. In such a scenario the effect on producer revenue is the same as in AMP groups because all members of the group consume the good even though he has sold only one instance of the good. The effect on the consumers’ values is altered in an OMP group, however. The non-paying members get to consume the good without paying the producer anything and with only the opportunity cost associated of spending time watching the film instead of doing something else and transaction costs associated with the existence of the group in the first place (both of which would still exist even if they had to pay). The paying member does bear the brunt of the cost, but he also derives social value either by increased esteem from his friends or simply from his enjoyment of the experience or something else. 

Enter the Internet

Friday, April 24, 2015

What’s So Bad About

Recently, several Indian companies left a zero rating arrangement called citing a need to defend net neutrality. While a strange move, the firms’ decisions are telling of the popular net neutrality zeitgeist. is one example of a broader type of practice called “zero rating.” This term refers to many different kinds of partnerships between mobile broadband providers and content producers. In the case of, Facebook is setting up agreements in developing countries in Africa, Asia, and South America where most people have never had any access to the Internet. In less developed countries, buying an allotment of mobile data is often the only way to get online. A zero rated plan means that users, can visit certain webpages without having the data counted against their allotment., for example, provides access to weather, health, and local news sites (as well as many others) and users can visit those sites without paying for the data used.

This sounds like a pretty good deal to me, but there is significant opposition to the practice of zero rating. Some governments, like Chile, have gone so far as to ban it altogether. The opposition comes from net neutrality fanatics. They accuse, and other zero rating arrangements, of unjustly favoring the content covered by the agreement and, thereby, creating barriers to the rest of the Internet.

These objections border on irrationality. First of all, there is nothing about that keeps users from visiting non-zero rated sites. They would simply have to pay for those sites like everyone else. If opponents are successful at eliminating zero-rating, consumers would have the same options that they would otherwise have but at a higher price. It is difficult to see how that position is pro-consumer. Furthermore, in the case of, most users now have no Internet access and so would be left with nothing if zero rating deals were outlawed. So what opponents to zero rating are really saying is that they would rather people have no Internet at all than the selection provided by

Not only do arrangements like not prevent consumers from visiting sites not within the free environment, the actors involved have an incentive to encourage leaving. Carriers only get paid if people click outside of the zero rated sites, and content providers will see an increase in demand as people who never had Internet access now demand content. Demanding some content from within the free offerings will inevitably lead to demand for other sites, not to mention the fact that many zero rated sites contain content which links to outside pages.

Many detractors have tried to nuance their objection by saying that they don’t mind what is doing, they just oppose the framing of it as charity because, they claim, it is merely an attempt by Facebook to make more money. Although is not exactly a cash cow for Facebook, there is a sense in which this point of view is exactly right. It is no coincidence that Facebook is one of the free sites in all of’s packages. But that is no reason to oppose it. There is no reason to think that a shrewd business practice must be bad for consumers. Indeed, voluntary transactions are, by their very nature, mutually beneficial. Other businesses are constantly instituting sales and giveaways to enhance their profits, yet there is no outcry against department stores every holiday weekend. No one seems to be proposing that no one be permitted to buy a car unless they are all the same price.

The paper thin nature of arguments against zero rating reveals the fact that opponents are not making a high-minded defense of consumer welfare. Instead, they are merely chanting the dogmas of net neutrality orthodoxy and demanding that all others to do the same. It is telling that the withdrawals of the Indian firms from simply said that they made the decision because they were “committed to net neutrality” or “standing up for net neutrality” without bothering to explain why net neutrality (or, at least, the zero rating subsection of it) is a good idea. If net neutrality means preventing willing actors from providing free Internet to some of the world’s poorest people, then maybe it is not worth defending. Under current conditions, zero rating deals like are unquestionably a win-win, and net neutrality zealots aren’t just looking this gift horse in the mouth, they are giving it a root canal.

This article was first published at

Wednesday, April 22, 2015

The Balancing Act of Patent Reform

Patent legislation is a balancing act the protection of legitimate patent holders and non-infringing operators while stopping malicious attacks from so-called “patent trolls.” Unfortunately, finding that balance is proving a herculean task; the current debate is controversial for good reason.

The Problem

Proponents of patent reform target their ire at patent trolls. These perfidious firms own patents that may be vague or otherwise of low quality. They then assert that small businesses, or others without large legal defense budgets, are infringing upon their patent and force them to pay a license fee or settlement lest they face an unaffordable lawsuit. The actual merits of the case are often irrelevant when even an innocent party cannot afford to prove it in court. A number of reforms have been proposed to combat trolls, and many have broad support. There are three issues, however, that are the main sources of contention.

1.       Who Are the Trolls?

Reform advocates are adamant that patent trolls are a huge problem raining fear and ruin on hardworking Americans. Others, however, question both the prevalence of actual trolls and the character of entities which are targeted as trolls. Common definitions of trolls call them non-practicing entities (NPEs) or patent assertion entities (PAEs) meaning that they hold patents but do not actually “practice” them (by producing the invention covered by the patent); instead, they merely “assert” their patents against others. But some have concerns over such a broad definition because not all NPEs or PAEs fit the description of a malicious troll. Several examples:
  • Some small inventors are technically skilled but don’t want to go into business to market their invention. They may choose to license the patented invention to others and live off the license fees.
  • Some patent holders may start out producing their invention, but the company may fail. The firm would still hold the patent, however, and can license it out to make the best of the situation. Other companies may buy the patents of failed companies in order to license and assert them. These businesses may provide incentives for innovation by ensuring that inventors have a market for their patent even if their business fails.
  • Universities are a third group that sometimes end up lumped in with trolls. Universities often invest in the creation of many new inventions, but they are not often engaged in producing them. Instead, they license the inventions to firms that will produce them.
The above examples would all be considered trolls, under many definitions, if they assert their patents against potential infringers. But all of them have reasonable claims to innocence from malicious litigation practices if they choose to sue.

2.      Fee Shifting

One of the most contentious proposed reforms is fee shifting. Some advocate a mandatory “loser pays” rule which would require entities judged to be trolls to pay the legal fees of their opponents if they lose their suit. The thinking is that since many trolls don’t really have a legitimate infringement case (or even a legitimate patent), they will not risk going to trial if they would have to pay the defendant’s legal fees when they lose. The measure would also make it more difficult for trolls to bully small businesses into a settlement because the alleged infringer, who knows he is in the right, can take the case to trial and then recover his legal fees.

Opponents are concerned, however, that fee shifting would discourage legitimate patent holders from enforcing their patents and might make them vulnerable to bullying from large, well equipped infringers. If a firm with a highly staffed and funded legal team infringes on a patent held by a smaller company or individual, the large firm will be able sit back and say “Go ahead and sue us. We will beat you and ruin your business by making you pay our fees.” So while mandatory fee shifting will discourage trolls, it will also discourage some responsible patent holders.

3.      AIA Reforms

Another interesting facet to the patent reform debate is that there was already significant reform just a few years ago. The American Invents Act (AIA) of 2011 was aimed at improving patent quality by allowing anyone to challenge any patent in what was meant to be a relatively quick and cheap proceeding before the newly created Patent Trial Appeals Board (PTAB) within the Patent Office itself. Here again, there are reasonable arguments from both sides.

Some point to the extremely high rate at which patents are invalidated in AIA proceedings as evidence that the system is overzealous and vulnerable to abuse. If a firm knows it has a good chance of invalidating a competitor’s patent, it can file for review of that patent to gain an advantage. Supporters of the AIA, however, hold that the high rate of invalidation by the PTAB early in its life only indicates that the lowest quality patents are being challenged first and the system is working.

Critics also assert that AIA proceedings provide an incentive to sell short on another company and then file for review of their patents knowing that the high success rate of such a challenge is sure to make the patent holder’s stock fall. Supporters respond by saying that such manipulation is already illegal under existing laws.

Another problem alleged by critics is the weakness of estoppel provisions. These are rules that prevent suits from being brought again and again with the same evidence. One of the AIA proceedings, covered business method review, has extremely weak estoppel rules. Petitioners are only estopped from using evidence actually presented at a previous trial.

This fact allows for abuses of the system by withholding evidence and filing again and again even if the PTAB finds the patent valid. Furthermore, the proceedings can be stacked so that patent that survives a standard reexamination can be challenged again under inter partes review. The result is a system vulnerable to abuse which could keep patents tied up in litigation even if they have previously been judged to be valid.

Patent reform is a balancing act with no easy answers. The unfortunate reality is that there are bad actors along with the good actors on both sides of the patent system; any tool that shifts the balance in one direction or the other empowers both the good and bad actors on that side. Eventual legislation on patent reform should exercise caution and humility in regulating such a complex and delicate arena.

This article was originally published at

Wednesday, March 25, 2015

3 Ways the “Open Internet Order” is Illegal

Now that the Federal Communication Commission’s (FCC) so-called Open Internet Order (OIO) has finally been released, we can begin to assess where we go from here. Besides the numerous reasons why the FCC’s decision is bad policy, it is also unconstitutional and a violation of the Telecommunications Act.

The Fifth Amendment

The OIO is a violation of the takings clause of the Fifth Amendment. The amendment concludes with the statement “nor shall private property be taken for public use without just compensation.” The legal history of this clause is long and complex, and the whole argument as it applies in this case is spelled out at length by Daniel Lyons. Here I will only summarize some basic principles that can be applied to the FCC’s Order. First, there is the rule, established by Loretto v. Teleprompter Manhattan CATV Corp, that a permanent occupation or regular use of private property by the government or the public is a per se taking requiring compensation. This is exactly the scenario that is created by the OIO. Internet service providers (ISPs) are now required to allow edge providers (e.g. Netflix, Amazon, Google) to use their networks without charge or face penalties for blocking and paid prioritization which the Order bans. The FCC and its allies have responded to this argument by claiming that use of a network is not physical occupation, and therefore not a per se taking, but this assertion is incorrect. The sending and receiving of information over networks takes place in physical space; the moving of content across a network with flows of electrons is not non-physical simply because the occupation is too small to see. The moving of information along a privately owned fiber or wire is not unlike the movement of people across privately owned land which the Supreme Court has already ruled falls under the Loretto rule. Furthermore, even if it does not require permanent occupation, the Order certainly requires ISPs to permit regular use of their networks, which still constitutes taking under Loretto.

The FCC acknowledges this difficulty in its Order (page 277 of the Order), but brushes them aside on the grounds that ISPs offer their services generally to the public. Since they voluntarily open their networks to occupation by customers, the Order argues, they cannot assert that regulation which prevents them from excluding certain customers is a per se takings. The Commission is correct in this argument; there is extensive legal precedent showing that regulators do have the power to enforce non-exclusion rules on firms which voluntarily offer their service to the public. The problem with the Commission’s argument is that the issue at play in the net neutrality debate, and the issue addressed by the Order, is not the relationship between last-mile ISPs and their customers but the relationship between last-mile ISPs and edge providers’ content. 

While it is true that exclusion of end users is not a per se right, exclusion of edge providers is. This important distinction can be illustrated with a non-Internet example: Walmart cannot engage in discrimination (for irrelevant reasons) against its customers; it has to let everyone shop there. Walmart does not, however, have to let everyone sell their products in its stores and it is allowed to charge a premium for things like prime shelf space. Therefore, it would be a per se taking for the government mandate that Walmart carry certain goods or all goods. In the same way, since the Open Internet Order requires ISPs to allow all edge providers to regularly use their networks, it is also a per se taking which requires just compensation. Since the OIO does not provide for such compensation, it is unconstitutional.

Mobile Broadband and Section 332

The FCC takes a new step in its Order by applying its common carrier reclassification to mobile broadband providers. Section 332 of the Telecommunications Act, however, prohibits private mobile radio services (PMRS), such as mobile broadband, from being regulated as common carriers. Indeed the D.C. Circuit in Verizon v. FCC, the case which struck down the FCC’s earlier attempt at net neutrality regulations, said “the treatment of mobile broadband providers as common carriers would violate Section 332.” 

Now the FCC is trying to get around that ruling (and its own previous rulings to the same effect) by arguing (page 15) that mobile broadband is the functional equivalent of a commercial mobile radio service (CMRS) which can be treated as a common carrier under section 332. The Commission’s argument makes a massive stretch by holding that because some Internet services, like VoIP, result in information being sent over the telephone network, they are therefore interconnected with that network which would allow them to be treated as CMRS. The illogical gymnastics of the FCC’s argument do nothing to change the clear fact that mobile broadband is PMRS; the provision of certain services that eventually link to the phone network does not mean that all mobile broadband is interconnected with that network. The stubbornness of reality means that the FCC’s classification of mobile broadband as a common carrier service is illegal.

Termination and Section 203

A stated goal of the Open Internet Order is to prevent paid prioritization of content from certain edge providers. The Commission wants to ban so-called “fast lanes” that would disadvantage smaller providers and entrepreneurs or allow ISPs to extort fees from edge providers in order to have their content delivered. The law, as Ford and Spiwak of the Phoenix Center demonstrate, makes the way the FCC goes about accomplishing this goal legally impossible. Section 203 of the Telecommunications Act requires that telecommunications firms submit positive tariffs to the FCC. This “tariff” is basically a list of how much the firm charges for each service it provides so that the FCC can judge whether the rates are too low (confiscatory) or too high (excessive). Since ISPs are now classified as telecommunications firms, they will be required to do this for their services.

The service provided by ISPs is obviously delivering to end users the content they request. But the Verizon decision indicated that ISPs also provide a service to edge providers known as termination. Termination is the service by which ISPs deliver the content of edge providers to that edge providers’ customers. For example, Comcast provides termination service to Amazon when it delivers Amazon content to Amazon’s customers who subscribe to Comcast Internet. Termination is basically the same thing as consumers’ Internet service but viewed from the edge providers’ perspective; delivery of YouTube videos to your house is as much a service to YouTube as it is to you.

Currently this system works fine because ISPs normally do not charge edge providers extortionate, or even any, fee for termination; they happily collect their revenue from consumers and go about their business. But with Title II reclassification, termination is now a telecommunications service, and, under section 203, ISPs must charge for it. Furthermore, the FCC cannot stop the creation of slow and fast lanes as “unreasonable discrimination” under section 202 because prioritized termination is a different service than regular termination. Discrimination cannot exist between unlike services. Ironically, the very scenario the FCC seeks to avoid is all but mandated by the reclassification.

The FCC seeks to escape this situation by forbearing from section 203 in the context of termination (page 241), but that is not legal. The Telecommunications Act allows for the Commission to forbear from certain provisions of the act or regulations if they are not necessary in order to serve the public interest. The FCC’s own argument, however, is that ISPs are “gatekeepers” for content going from edge providers to the ISPs’ customers, and they must be regulated to ensure the public interest. The D.C. Circuit agreed saying in Verizon that ISPs were “terminating monopolies.” As Ford and Spiwak observe, in all previous forbearance cases competition was cited as justification for why forbearance was consistent with the public interest. Since the FCC and the courts have already said that ISPs have a monopoly on termination, they have no legal basis for forbearing from section 203 for termination service. The FCC is aware of this problem but ignores it merely saying that they “reject the argument” (page 197).

Despite widespread enthusiasm about the FCC’s decision, the Open Internet Order sets aside both sound policymaking and the rule of law. The Order should not survive in court.

The article was originally published in two parts on Part 1 Part 2