Paper on network effects in the video game market

In the course we have been recently discusing network effects. Indeed, network effects and positive feedback loops have received a great deal of attention, academically and otherwise. In a market with network effects, competition among multiple incompatible systems is intense, because a small, initial advantage confers a larger advantage in the future. Many theoretical papers suggest various competitive strategies in a market with strong indirect network effects, but little work has been done on what strategies are most effective in each phase of the product cycle. The paper I linked to above
analyzes two sides of the U.S. video game market, hardware adoption by consumers and software provision
by game makers, and estimates the elasticities of adoption with respect to console price and software variety.
It found that the relative size of the elasticities of hardware demand differs over the product cycle: When
a console is introduced, hardware demand is quite elastic with respect to price, but much less elastic with
respect to software variety. As the console becomes mature, the price elasticity declines substantially, but
the elasticity with respect to software variety increases substantially. The estimation results suggest that,
while a sufficiently large set of software may be necessary to launch a system, a platform provider should use
penetration pricing to encourage adoption at the outset (i.e., a lower price-cost margin). Once the platform
provider succeeds in establishing an installed base, it can expand the installed base, and thus the profitability
of the platform, by encouraging software entry. A wider variety of software is crucial for attracting later
adopters to the platform. The areas of analysis covered by the paper helped to shed light on network effects in the video game market, but neglected to fully discuss the varying incentives of platform providers . The different supply strategies used by platform providers seem to have an effect on console adoption by users. It would seem that some systems gain popularity by having an initial scarcity; this scarcity causes the system to have the illusion of a network effect, which leads to the actual network effect of console adoption.

http://www.nytimes.com/2007/03/30/science/30sharks.html?_r=1&ref=science&oref=login

The food chain is a perfect example of a natural graph. This dependency network graphs different species as nodes, and links them to other species/nodes that they consume. However, what would happen if you increased the “flow” along one of these edges? How would this effect the rest of the chain? In a recent study, researchers are concluding that the decrease in shark population has caused scallop beds in North Carolina to be practically wiped out. The absence of these large predators is mostly due to overfishing and an increased demand for meats and fins, something that humans can be entirely attributed to. This has all caused sharp declines in the shark population, which in turn has caused an increase in all the creatures that sharks consume. These creatures mainly consist of smaller sharks, skates and rays. The increase in these animals has in turn caused a decrease in their prey. In this article, the decrease in sharks led to an increase in cownose rays, which has led to the nearly wiped out scallop beds in North Carolina. This entire process demonstrates the idea of trophic cascades. This occurs when the predators in a food chain suppress the majority of their prey, and thus, free the next lower level in the chain from predation. The abundance of cownose rays during their southward migration along the Eastern seaboard resulted in the depleted scallop beds.
By drastically decreasing one node in this graph, you get an accordion effect on the rest of the graph. The level just below it will increase and the level just below that one will decrease, and so on and so forth. Researchers are trying to determine how long this effect stretches down the food chain, and how meaningful it actually is. It is interesting to see how the decrease of an extremely large predator, which is high up on the food chain, can cause so many problems for the rest of the ecosystem. This trophic cascade could stretch even further than the scallop beds, if other animals depend on those scallops for survival. The entire food web is affected because of the drastic impact that the decrease of large sharks have on the ecosystem.

http://www.portaliraq.com/news/Iraq+Museum+International+finds+cheating+on+eBay__1111434.html

This article discusses a “sniper trap” on eBay. The idea is that the seller and a friend attempt to raise the bids on an eBay auction. As the auction is closing, the seller’s friend poses a fairly high bid, thereby triggering the secret bids of genuine bidders. These secret bids are revealed and the seller cancels the fake bid. Now genuine bidders will attempt to outbid the other secret bids. In this way, the seller increases the competing bids. This sniper trap has been successful in generating higher last minute bids.

In class we have discussed various types of auctions. Most of these auctions are fair auctions, and using this assumption we have been able to explore certain rules and trends of auctions. For a fair second price sealed bid auction it has been shown that a bidder should bid his valuation. In eBay, the auction is a variation of the second price sealed bid auction , so people bid up to their true valuation. It seems odd that eBay would allow a seller to participate in his own auction in this way. If the seller is allowed to cancel bids at will, then he has the ability to drive the prize higher by having people place phony bids. If these are outbid then the seller gains, and if not they can simply cancel the bid. It seems that a more fair auction would prevent this participation or at least prevent it in the final hectic moments of bidding.

An interesting question is to see what the bidders should do if they suspect that the seller will employ this technique. Although in a fair auction the bidder should bid his/her valuation, it would be more advantageous to set a secret bid that is lower in this case. In this situation the highest bidder has a lower secret bid. When this bid is revealed, the genuine bidders will attempt to outbid this secret bid. Then the bidder can bid his true valuation and hope that his bid is higher than the price war caused by the unfair bid. This particular technique would most likely not be advantageous if a single bidder employed it. However, if the sniper trap becomes more widely used, more bidders may attempt this technique, thereby shifting the dominant strategy from bidding true valuations to bidding lower secret bids.

The canonical example of political Google-bombing was perpetrated in October of 2003, when a large group of bloggers linked the terms “ miserable failure” to George W. Bush’s official White House site biography. Indeed, the internet is one series of tubes that is rife with such politically motivated chicanery. One example, though perhaps not as notorious as the “miserable failure” bombing, was highlighted in an International Herald Tribune article.

Chris Bowers, a contributor to the progressive blog MyDD, requested during the 2006 election cycle that whenever a poster mentioned a Republican candidate, that official’s name was to be linked to a damaging news article or op-ed that Bowers had found regarding that candidate. For instance, Senator John Kyl (R-Arizona) was given the distinct honor of being linked to this flattering Phoenix New Times article, entitled “Stealth Zealot”. The result of this particular link bomb can still be seen on the first page of Google search results. Though while these stunts are often amusing, their efficacy is often dubious as those of different political-persuasions may not appreciate a political Google-bomb’s levity and many may not appreciate what they consider search-result spamming. When link bombing was first used, Google itself received criticism as those unfamiliar with Google’s page-ranking system thought Google itself endorsed the opinions expressed by the bombers. Many also predicted that with the blooming blogosphere, the quality of Google’s search results would suffer, being too easily influenced by relatively few individuals.

Despite the criticism, and often hysterical portents of doom, Google insisted that it would maintain its objectivity and keep its ranking system relatively untouched. However, in January, 2007, Google waffled and changed its search algorithm in an attempt to prevent instances of link bombing by massive link farms. Some rumors on the internets have speculated that the revised algorithm only incorporates a blacklist mechanism rather than a fundamentally altered page-ranking system. Those who support this claim cite the fact that prominent Google-bombs such as the “ French military victories” bomb still exist, whereas other well-publicized Google-bombs have been eliminated, and more have been eliminated.

As beefcake points out in a recent post, there is significant need for effective web page filtering. Due to the web’s monolithic size and the dynamic nature of its content, effective filtering becomes very difficult. In class, we focused on exploiting the hyperlink structure of the web to rank pages relevant to a query. Though this is a powerful, proven method used by Google and other search engines, it confines the user to using queries to explore the web. While this helps users locate relevant items, it is not particularly useful for discovery of things that they aren’t yet aware of but may find interesting.

A few sites are now offering a clever, more personalized search approach. These systems take advantage of the power of online social networking to filter the web’s content and help users discover new and exciting things. Digg, for example, relies on users to submit articles they find on the web. These stories are then voted on by other users, which helps determine whether or not they get promoted to the main pages. In addition, users may befriend others to view the stories they have submitted or voted on. In her article, “Social Networks and Social Information Filtering on Digg” Kristina Lerman examines how these sites have transformed the web into a participatory medium. These sites enable users to evaluate content and share their opinions with others across the web. This represents a shift in the web usage paradigm. Rather than merely querying the body of information on web pages, users are now assuming an active role in the creation, evaluation, and distribution of online content.

In an effort to examine the effectiveness of Digg as a recommendation system, Lerman examines the effects of using social networking for content filtering. She determined that users are more likely to signify that they like, or “digg,” stories their friends submit. Also, user’s statistically digg stories that their friends also digg. While these properties indicate reasonable social filtering, Digg’s system is not perfect. The characteristics of Digg’s social network can result in overrepresentation of a relatively small minority. This is because of the network effects experienced by the top users. Those with the largest network of friends are ideally positioned for achieving a large numbers of diggs on their future posts. Also interesting to note is the tendency of some users to add large numbers of friends. This could lead to some herding effects seen with information cascade. Users may be more inclined digg content merely because it is popular among their friends.

Del.icio.us is another interesting social networking system which relies on its body of users to bookmark content across the web. In addition, it allows users to tag the content using keywords. This adds useful metadata which categorizes the content and allows users to browse information more efficiently. This accomplishes the filtering suggested by beefcake, and helps avoid the issue of the web site creator intentionally misrepresenting the information for personal gain by delegating this responsibility to a larger, hopefully unbiased user base. Though it is far from perfect, social filtering is a promising technique which offers some unique advantages over traditional query-based systems.

http://www.usatoday.com/money/markets/us/2005-11-10-pink-sheet-cover-usat_x.htm

The article describes the techniques scammers use to profit from the lightly regulated pink sheets stock market. This stock market is known for trading stocks that don’t meet the requirements to be on a larger stock exchanges such as the NYSE. Most of these stocks, which trade at low prices and volumes, are “shell companies,” which means they don’t have any real assets or operations. Scammers use a “pump-and-dump” scheme to advertise the penny stock to uninformed investor. Some of the methods scammers use to pump the stocks include questionable correspondences, fake research reports, reverse mergers, and multimedia assaults. We will discuss how the pink sheet stock market is a perfect framework for a simple one-way information cascade and how the internet can be used to maximize the probability of a successful pump scheme.

The pink sheet market provides an excellent framework for information cascades because of the nature of the stocks and investors. Compared to reputable stock markets, the pink sheet market contains stocks of low volumes. When a few investors buy into a stock, it can raise the share trading volume by hundreds of percent versus the small percentage in other stock markets. The drastic volume changes can give strong buy signals to pink sheet investors. Most of these pink-sheet investors are less informed than ones who trade on the reputable stock markets and are more likely to make investments decisions based on other people’s actions rather than their own information. By ignoring outside information, investors make the cascade more likely to accelerate rather than stop.  In short, they buy stocks more mechanically based on the other people’s actions, similar to the simple one-way cascade effect we discussed in class.

All the scammer needs to do is get a few unsuspecting investors to buy the stock, which will send “buy” signals to other buyers. Since the internet allows for the mass dissemination of information, it enhances the scammer’s opportunity to reach uninformed investors. Old forms of advertising like fax and voice mails are not nearly as useful as spamming and posting in investing communities. Spamming allows scammers to reach millions of people, and chances are that a few people will believe the sensationalistic email stating the stock that will go up 500% in the next week. Investing communities are another great place for scammers to promote their stocks, pretending that they have some inside information about the stock. On top of all of this, financial sites such as Yahoo finance list these stocks, which appear more legitimate than they really are. The pink sheet stock market needs more oversight to help protect unsuspecting investors from scams.

We are all plagued by spam in our in-boxes and telemarketers calling us at dinner. But, for the most part, cell phones have been free from mass marketing strategies. A patent has been issued (http://www.patentstorm.us/patents/5852775-description.html) for a system that may change this. The cellular telephone advertising system would play ads of up to 30 seconds at the beginning of every conversation originating from a cellphone. When a customer places a call, the system will look at demographic information about the customer as well as information about the cell they are connected to (providing geographic information) and play an advertisement before connecting the call. Theories similar to those used in key-word based advertising can be used to select ads which customers will receive. Knowing the customer’s location allows advertising for close by businesses, which they are more likely to visit then distant ones. Also, demographic information (i.e. sex, ethnicity, etc…), allows targeted advertising (for example a person is more likely to visit a local restaurant if they serve their favorite food).

So if you place a call while driving by a McDonald’s, you may very well be forced to listen to an ad about the latest additions to the dollar menu before being able to connect. This technology raises concerns about customer’s privacy, as well as forcing people to listen to an ad before being able to place a call (which may be potentially urgent). There are some current examples of being forced to listen to advertisements, like commercials on TV or internet video cites, but this seems to take it to another level. If this system is actually ever implemented, people will be forced to listen to advertising before they can use a phone.

Closely connected to the concept of information cascades is the self-organizing system. Such a system will organize itself without any centralized hierarchal structure, but rather through many automatic decentralized processes.

One particularly striking, if morbid, example of a self-organizing system is an ant mill: first described by naturalist William Beebe, an ant mill occurs when a large number of drones are separated from the colony. Unable to find their way back, the drones walked in a large circle (1200 feet in circumference) for days, until they all keeled over dead. They were all obeying a simple rule while lost – follow the ant in front of you.

These ant mills are actually fairly common, and are often stopped by a single ant or small group of ants straying from the course accidentally, thus inadvertently saving their entire colony. An ant mill is a clear example of an information cascade, both as ants enter and leave the mill, and of a self-organizing system.

Another example is certain species of firefly, which will perfectly synchronize their lights within a certain degree of locality. Again, this happens without any centralized organizational structure, and may be due to a process very much like the one we’ve been discussing in class, although it’s not fully understood how the bugs synchronize.

In a paper written by Thomas D. Seeley at our very own Cornell University’s department of Neurobiology, he defines a system as self-organized whenever it contains a large number of subunits which lack the wherewithal to implement centralized control. Note that this perfectly describes a college student body, in which many examples of information cascading may be seen vis a vis fashion, course choices, and so forth. However, a student body doesn’t really achieve any communal goal, so it’s hard to classify it as a self organizing system. Maybe if we weren’t so lazy.

Thomas Seeley paper: http://www.biolbull.org/cgi/reprint/202/3/314.pdf

Circumnavigating the blogosphere today was an article entitled “How to Ace Your Finals Without Studying”. Naturally, as any student would, I read the article hoping to find some divine secret to enable me to ace tests without studying.

The premise of the article is that it is much more efficient to study in a way that forms relationships within the material in order to better understand it. Forming spiderwebs, as the author put it, enables us to recall information faster and better than the opposite way to store information, compartmentalizing. His analogy:

Compartmentalized learning is an exercise in insanity. A comparable strategy would be if the users of the web didn’t hyperlink anything. So to find any information you just had to keep typing addresses into your browser, hoping that it would pop up. Studying for these learners is akin to setting up thousands of domain names that all lead to the same information, so that you will hopefully get to the right place by just guessing enough.

To me, this is a great example of the importance of relationships and networks in something outside the internet and inter-personal connections. And while this all sounds great, implementing it could be a struggle. I can’t imaging drawing out a web of the terms involved in Psych 205… it would take a huge piece of paper and many hours to accomplish, and that’s only if you could appropriately link each term to the others. However, discovering the links among data is part of the learning process, the article says. Asking how one piece of information relates to other pieces is part of the process in forcing your brain to remember the webs so that you can recall the information later.

After I read the article I spent a few minutes looking for software that could do this type of thing effectively. As I review chapters in a textbook, it would be great to enter them in a system and easily form the connections to other pieces of data. I couldn’t really come up with much. Microsoft Visio does something along these lines, but it’s convoluted and expensive. Bubbl.us does the job fairly well, but it’s web-based and fairly light on the features. If anyone knows of better brainstorming/diagramming software, drop me a line.

Movie Stars, Big Budgets, and Wide Releases: Empirical Analysis of the Blockbuster Strategy, a paper by Arthur De Vany and W. David Walls

Does the blockbuster strategy really make people herd to the movies? This is the theory that strong advertising, big-name stars, and huge opening nights are what make movies successful. In this case, the early choosers would be the ones making all the decisions. De Vany and Walls argue that this non-informative information cascade (in which later movie-goers are choosing not based on information but based on pure mimicking), would mean that potentially horrible movies with huge opening nights would always dominate in popularity. They prefer an informative cascade, in which rather than a blind following, there is a chain of communication about the movie’s quality that leads to higher numbers in the theaters.

Of course before a movie is released, there is little word-of-mouth spread of opinion. So opening night does often follow this blockbuster strategy. This non-informative information cascade only counts when it is the sole fact that other people are going that results in additional watchers, something that can be triggered for example by watching box-office revenues alone. Unfortunately for the superior-budget-but-not-so-superior-film producers, movie-goers are not lemmings and a hit on opening night does not signify continued success. The cascade becomes informed after the kick-off when personal judgments start spreading and people’s decisions to see the movie are now based on opinions rather than fad, and quality begins to matter.

De Vany and Walls attempt to analyze the actual historical balance between the two types of cascades, one based on others’ actions, one on their opinions. Through the use of statistics, they examined the effects of budgets, stars, openings, genres, and rating.

As an interesting side note, the paper identified the stars that seem to most influence the revenue of a film, either positively or negatively. (Keep in mind that the sample consisted of movies that came out between 1986 and 1995.) Good people to have in or working on your film back then: Tom Hanks, Oliver Stone, Michelle Pfeiffer, Sandra Bullock, Jim Carrey, Jodie Foster, Steven Spielberg, Barbra Streisand, and Robin Williams. It was apparently not a great idea however to have Val Kilmer, Winona Ryder, or Sharon Stone.

We have only begun the discussion on information cascades, but I think this serves as a great example of the phenomenon. It encompasses two types, the non-informative and the informative, and shows that while some people choose to accept a choice based on simply following-the-leader, it is the latter acceptance decision based on positive word-of-mouth that makes a difference in a movie’s true success.