Wired News copy chief Tony Long recently posted an interesting article entitled Parking Goes to  Highest Bidder.  The article deals with a startup company in Cambridge, Mass.  named SpotScout.  SpotScout aims to hire a group of “spot scouts” to keep a sort of inventory of local parking garages.  Price and availability information on participating garages is uploaded in real time to the website.  The information is then made available to all subscribers on any web enabled device.  As an interesting twist SpotScout also has a feature to allow private residences to register spots in their driveways to sell.  It also plans to have “Spot Casters” to search for open public parking and also may allow drivers to buy and sell information on when they will be leaving their public spots.  Public spots cannot however be reserved as garage spaces can, you are simply paying for proprietary information on location and time of openings.

This system seems that it could be very interesting from a price dynamics point of view as it represents an amalgamation of a lot of concepts from class.  In examining the underlying market networks it is first important to point out that the article title is a slight misnomer.  The parkers do not actually bid on prices.  In fact the set up is almost more of a dutch auction except there are many sellers and sort of an unknown, variable, number of buyers at any time.  The major twists are that people not directly involved in the auction (i.e. parkers not using SpotScout) are also competing for the same spaces, they just don’t have knowledge of all of the sellers.  The other twist is that the garages have web consoles that allow them to adjust the price of spaces listed on SpotScout in real-time (not all the spaces in one garage are necessarily listed).

This pricing system also incorporates the use of a trader, namely SpotScout.  The premise is that most people are not familiar with all of the local garages and even if they are, do not want to have to drive into all of them to see if there are available spaces.  In this case the trader makes money by taking a fixed portion of all transactions (registration for garages and parkers is free).    It would be interesting to see how SpotScout picks the transaction fee because unlike the trading networks we saw in class, the buyers are not required to even use the trader, they can go directly to the seller.  All in all it will be interesting to see how this company will affect parking prices in Cambridge.

Using the Web to Get Your Boss to Pay More

Information is power. Employees now have the convenience of comparing their current salary to other people’s salaries who have similar job titles by using sites such as PayScale.com and/or Salary.com. These web sites aim to level the playing field by allowing people to access detailed information on salaries, giving them more leverage while negotiating with their bosses for a pay increase.

This idea ties into the network exchange theory discussed in class; initially, if there are only two negotiators, A & B, the split it usually equal and each person gets about half. However, if an outside source like another offer or a website, like PayScale.com, is available to person A then person A can have advantage over person B. In the working world, bosses and their subordinates cannot be modeled as equals such as persons A and B, and so these salary websites serve to give employees more power in negotiations to balance the relationship.

The article also discusses potential problems in using this data. There was a survey conducted that found 96% of human resources departments did not trust the data from PayScale.com since it is user-reported. People can misreport information if they don’t understand what types of benefits or bonuses they are getting. In addition, the information on different jobs is completely dependent on the number of people who decide to contribute and therefore, the data is most accurate when there is a lot of data. These sites still need to improve the range of information they have on different jobs. Also, in asking a boss for a raise, an online source with salaries does not always suffice because other things need to be taken into account such as the company’s current financial status and historical turnover rate.

http://xxx.lanl.gov/abs/cond-mat/0409609
(click on the PDF link at the bottom for full text)

Having played soccer for a majority of my life, I’d like to think I know everything there is to know about the sport. Though this may be true about the rules and strategies of the game, after the reading the Palacios-Huerta paper I found a new lens with which to look through when analyzing soccer. This academic view that utilizes the world’s most popular sport for game theory applications is also being used to illustrate some interesting relationships found in social networks, namely those amongst players from various clubs. In their paper, Roberto Onody and Paulo Castro study what they call, the “Brazilian soccer network.”

This fairly simple network is composed of a bipartite network that contains two types of vertices; one set is created from the 127 different Brazilian soccer clubs and the other set is created from the 13,411 soccer players who have played for those clubs between the years 1971 and 2002. Much like the networks studied in class, edges are formed based on a social connection. In this network, “whenever a soccer player has been employed by a certain club” they are connected by an edge. Throughout their short but very analytical paper, these two researchers explore a number of relationships and then apply mathematical models to their results. For instance, they found that the probability, P(N), that a player has worked for N clubs follows an exponential decay model, a result of which shows that “it is 190 times more probable to find someone who has played for only two clubs than for eight clubs.”

Subsequently in the paper, the researchers create a “unipartite network composed exclusively by the soccer players. If two players were at the same team at the same time, then they will be connected by an edge.” From this network, they analyze a number of variables that are relevant to our class including the number of edges formed and the average shortest path length between players. With regards to the average shortest path length, they found that within the Brazilian soccer player (BSP) network the value was 3.29. “In analogy with social networks, we can say that there is 3.29 degrees of separation between the Brazilian soccer players, or, in other words, the BSP network is a small-world.”

Onody and Castro leave the reader with some interesting thoughts about their results, postulating as to why such results were found. For example, they think that the “player’s professional life is turning longer and or/ the players transfer rate between teams is growing up.” Lastly, they hope to stimulate further research in the subject, asking whether their results would “hold for soccer players from different countries or, perhaps, different sports.”

In the paper, there were a few variables and relationships that were discussed that I would appreciate knowing a bit more about, namely the “clustering coefficient” and the “assortativity coefficient”. Though the clustering coefficient has been touched upon, perhaps if there is enough time in the course we could review these “relevant parameters in social networks.”

The particular part of Shannon Applecline’s “Social Gaming Interactions, Part Two: Competition” which pertains most to the material in this class is what she describes as “the most exciting competition,” the economic competition involving auctions. Especially of interest is the variations on common auction types
which these games display, and how auctions traditionally held in reality are represented in digital form. Take the Dutch auction for example. This is easily digitally modeled in Merchants of Amsterdam with a clock which counts back and reduces the price simultaneously. First-degree price discrimination is displayed in Queen’s Necklace, which has a turn-based Dutch auction where players buy a small quantity of some item when they can afford it, and in the next round, prices drop and the game continues.

Interesting variations of auctions are sometimes done at the expense of the bidder. The reason for this is to curb devaluation of currency, which is quite rampant in online games. Fist of The Dragonstones uses a sealed-bid auction where all bids are spent and only the winner gets the item. New England displays a typical English auction with a twist, players only bid in certain rounds and have to pay in order to bid in earlier bidding rounds. Another solution to the devaluation of currency is displayed in Res Publica, where a seller offers an item, and bidders offer other items in trade or a bidder requests an item and sellers offer items. As a creative medium, games should continue to offer new and interesting variations of commonly used auction techniques.

http://swicki.eurekster.com/ 

Swicki is a search engine which strives to provide customized results that are more relevant than a general search on a regular search engine.  This is done by creating a Swicki, a customized directory with preferences that are specified by the creator.  A directory is initialized by filling out a survey of preferences, thereafter, search results are updated based on click patterns of all users (within a Swicki) on results of all searches to improve results and update parameters for more relevance on future searches.    Users can pick a directory to search in and results will be filtered to give links which are more relevant to target group specified by the Swicki.  Users of a specific Swicki can be referred to as a community consisting of users with similar interests.   

The site is the core of a large network of Swicki users, which can be broken down into sub-networks consisting of users with similar interests, “like-minded users”.  Users may or may not know each other.  Since members of the communities on Swicki are private and anonymous, it is unlikely that users of the same Swicki directory have a relationship with each other; however each member of a community has a dependence on the other users as they all affect the search results.  If we were to attempt to draw a graph to depict this network, with a square node representing websites that users click, circular node as users and edges showing that a user has clicked a website; we would have a graph of many circles with arrows towards different squares.  A community would be identified by a group of circles connected to similar squares; while the users are dependent on each other they do not interact and are not connected to each other on the graph!  Is there a better way to represent the communities? 

Over the past decade, the number of blogs on the internet, and the connections between those blogs, have been growing at an enormous rate. There are blogs that discuss technology, blogs that discuss personal lives, and numerous blogs that discuss politics and news. In Issue Publics on the Web: Applying Network Theory to the War Blogosphere , a group at the University of Texas at Austin discuss the makeup and interconnections of the portion of the blogosphere concerned with the Iraq war. Their primary object is to observe and characterize the differences between the liberal and conservative blogs in the sphere of discussion.

One particular measure of each blog that the researchers examined was the number of incoming links to each blog. Larger numbers of incoming links was taken as an indication that a given blog held a powerful position in the network, as it was getting more attention than its peers. This connects nicely to the intuition about node power in graphs that we have been discussing in class. As we have seen, nodes that are in a more central position tend to have more power in the graph, and a larger number of incoming links indicates that a blog is likely to be relatively central in a network (especially using betweenness as the measure of centrality).

The group was able to determine the correlation between a number of different factors (number of authors, gender of the authors, types of posts, and number of outgoing links) and the number of incoming links to a given blog. Outgoing links had the largest correlation (again, indicating that blogs that are central clearing houses for information tend to have more power and influence in the graph), followed by the types of posts made.

All in all, this seems to be an interesting analysis of a very restricted subgraph of the larger internet. Because of its limited scope of study, the graph is both easier to understand and analyze, and could be a useful predictor for the behavior of graphs of blogs in general.

“Tag everything!” It’s the rallying cry of Web 2.0. Bookmarks (Del.icio.us), pictures (Flickr) or videos (YouTube), it can all be tagged.

Read the rest of this entry »

Social Networking’s Next Phase

This article discusses the idea of using social networking for customers of a business.  Cisco Systems and other companies are exploring the idea of setting up a social networking service similar to Facebook or MySpace for companies to allow their clients to interact with one another.  This presents an interesting idea since most people on normal networking sites do not share a lot of common interests with others on the site (with the exception of their close group of friends).  In a site set up for customers, most, if not all, of the users will have several common interests in what that company is doing and how the company can help further their own goals.

This also allows for the idea for individuals to now set up their own social networks.  Individuals could custom-make a social network for a group or idea that they are interested in.  This would more closely connect the individuals who are linked on that particular site.  In settings such as MySpace, if users are not actually linked in an offline environment, then it seems that it would be rare for users to become “friends” if they share no interests.  However, in a niche social networking site, these users all share some interest, so it seems like they would be more willing to interact.  This means that triadic closure should be more prevalent in a site such as this.  If users do not already share some offline bond, then they might be more willing to meet others in a setting like this.  For a site such as MySpace, users that are friends online seem to most likely be friends offline.  If triadic closure does occur, it may not have a direct correlation to an online relationship; rather, the new friendship may have be formed when an acquaintance or friendship is made offline (and then it is reflected online).

One caveat for a company such as Cisco is how much attention they can garner for their new business.  Well established social networking sites almost have a monopoly on users; therefore, without advertising or some other way to spread the news about their new service, it could go virtually unnoticed.  The way that popularity is garnered for a project like this is a social networking project in itself.  Presumably users will spread news of this service if they like it and want others to use this.  This could be done by word-of-mouth, or by other means such as a weblink from a user’s own home page (in which popularity for the service is then determined on how many views this website receives).  Assuming that this spread does occur, this presents a very interesting situation.  Users are now presented with the opportunity to “fully” customize their web experience instead of having to conform to MySpace or Facebook restrictions.  This could push these established giants into a battle where they will have to change the way they run their sites.  No matter which site(s) a user chooses, it will be interesting to see how the growth in different social networking groups can affect the overall landscape of the web in years to come.  Will social networking take a more dominant role in how the web is structured, or will things remain constant (or take another direction) in coming years?  This should be reflected in how well these niche social groups can prosper in the current web environment.

http://webspace.ship.edu/cgboer/selfdefense.html

In his paper Self-Defense: Balance Theory, Dissonance Theory, and Transactional Analysis, Dr. C. George Boeree (professor in the psychology dept at Shippensburg U.) takes the concept of structural balance theory that we discussed in class, and places it in its larger social psychological context.  Boeree argues that Heider’s Structural Balance Theory, Festinger’s Dissonance Theory, and Eric Berne’s Transactional Analysis theory are all mechanisms of self-defense, employed by people to achieve a more balanced internal state.

In class we learned about Structural Balance Theory and applied it to social networks as well as to graph theory. We learned that in order for a labeled, complete graph to be balanced, either all nodes must be linked positively (all individuals are on friendly terms) or the network must be divisible into two factions with all nodes in faction one linked positively and all cross links between the two factions negative. In general, we noted that the product of all signs must be positive (ie- 3+ or 2- and 1+, etc.)

 Boeree’s article looks at the different “self-defensive” tactics people employ in unbalanced situations. He first touches on Structural Balance Theory,  looking at what occurs psychologically  in order for balance to be achieved in social networks. Heider noted that unbalanced situations cause stress to the individuals involved, and this stress promotes change- perhaps by convincing ourselves that we like something/someone more than we actually do. Boeree then cites Festinger’s dissonance theory, which states that when our actions contradict our self-concept, it produces an alarming state of dissonance (unbalance- emotions such as guilt), which in turn produces an action (such as inadequate justification to explain our actions). Finally,  Boeree notes Eric Berne’s theory of Transactional Analysis. This theory illustrates the ego as having 3 components (parent, adult, child). By placing 2 egos side by side, a social network is produced that can illustrate different “transactions” between individuals. Boeree notes that individuals engage in game-like transactions as means of self-defense. He gives several examples of this in his paper, and goes on to argue that dissonance in social settings (not just individual dissonance) can be remedied by these transactional mechanisms of defense.

I’ve always wondered how the various coffee shops in Collegetown have determined their prices. Using what we’ve learned in class, can we model the Collegetown coffee “scene” with some auction and game-theoretic tools? And will this model accord with what we are observing in practice?

Setup:

There are four major coffee shops in Collegetown: Collegetown Bagels (CTB), Starbucks (SB), Wilson Farms (WF) and Eat Desserts First (EDF). Many more places sell coffee, but these are my usual destinations. My independent private values (IPV) for where I purchase coffee are a combination of the price of coffee and the distance I must go out of my way to buy coffee (really, this is just to say that more affects my preferences than just price):

IPV = 1 / ( price of coffee + distance )

My preferences can be thought of as a ranking of coffee shops, with my most preferred coffee shop(s) having the largest IPV. Currently, I prefer WF coffee because it is the least expensive and the least out of my way. I visit my preferred coffee shop(s) with a certain probability p > .7, because I do wonder what else is out there (1-p) of the time. However, since I’m trying to curb my caffeine addiction, I will only drink a single cup of coffee a day. If I have more than one preferred coffee shop, I will chose a destination from my preferred set with equal probability.

I am a very picky consumer and I like to think that all coffee shops exist to please me. More specifically, assuming all the coffee shops equally want to attract me to purchase their coffee, how would they set their prices?

Networked Coffee

We can think of the four coffee shops, CTB, WF, EDF and SB as Seller nodes in a bipartite graph with a single buyer, me. Currently, the preferred seller graph has a single edge between the buyer (me) and WF, as I prefer their coffee and no others. The three other sellers, assuming they cannot move locations, will try to increase my IPV for their coffee by lowering their prices. If their coffee’s IPV is equal to the maximum IPV, I will be equally happy (indifferent) between buying coffee at their establishment and some other preferred shops, and they will get more of my business than if I did not prefer them.

A perfect matching will be a set of market clearing prices. As in class and previous homework, we may create 3 fake “buyers”, for the purpose of matching each seller with a buyer.

A subtle point in this network is that the smaller the set of preferred shops, the greater the benefit for the preferred sellers. Since each preferred seller has an equal probability of being selected, the fewer the options, the greater patronage each seller will receive. Coffee shops will be motivated to lower their prices as far as possible because we will assume the patronage they will receive for being the single preferred seller is more important than the difference between their lowest prices.

Bean Bickering

On an individual level, each seller is motivated to lower their prices until they are in the preferred seller set. However, the coffee shops cannot arbitrarily lower their prices indefinitely. As in class, where houses were not sold for less than $0, we should cap the lowest price a particular coffee shop can charge. Since coffee proprietors are a crafty breed, they will only lower prices to their private “true” cost to produce the coffee, ensuring that they will never lose money on a sale. This is like an auction where sellers will sell their coffee at a price equivalent to the second lowest bid (which causes the second lowest seller to drop out), and therefore they will benefit from shading their true cost to produce coffee. Price wars can occur over many rounds and many different price postings. The winner’s dilemma applies here: if you “win” my business, you are valuing my business at a price below any of your competitors’ prices.

Secret Sourcing

What complicates matters further is that each seller now has a minimum price equal to their true sourcing cost – also an independent, private value. They are shading their prices, and they benefit when no one else knows their cost. My freedom in the coffee buying market closely resembles a network exchange model with perfect competition, where price wars between coffee shops may result in 0-profit sales at sourcing costs. With the slippery slope of network-exchange price wars, coffee shops must think carefully before lowering their prices. While lowering their price may add them to the preferred seller graph, it may also initiate a price war which they will lose, where the perfectly competitive price charged is lower than their true cost. If they lose the price war, their situation is worse than their intitially higher price at (1-p) frequency.

Since sourcing costs are secret IPV’s, the risk of initiating a network-exchange 0-profit price war is great. Much like the Prisoner’s Dilemma, all coffee shops are worse off if everyone lowers prices, while everyone is better off if no one lowers prices. If only one coffee shop lowers prices (becomes a preferred seller), that shop is maximally better off.

Stop, Drop, or Roll

It seems that in order to attract my business, coffee shops must evaluate the risks of lowering prices and losing money by engaging in a price war versus the risks of being in the not-preferred set. If coffee shops find the risks small enough, they will lower their prices incrementally, shading their true value to increase profit – going no lower than their true coffee-sourcing costs. They will either end up in a preferred seller graph, or hit their true value and remain in (1-p)-land. Remember, there is no guarantee that sourcing costs are unique. If coffee shops find the risks too high, they will hold prices relatively constant. Lowering prices seems to be “a roll of the dice” for coffee shops. In the market as I have described it, coffee shops are at the mercy of the consumer (me) unless they collude to fix prices.

Personally, this analysis seems to indicate that one should expect relatively stable coffee prices proportional to private costs and current trade volumes. Prices will not change much, for fear of fierce competition. Instead, we might expect proprietors to try and attract my business through non-price-lowering incentives, that would not trigger a price war. Indeed, this is supported by CTB’s new and unheard-of “Coffee Hour” discount!

Evidence of network markets at work!

Some questions for further thought:

1. Clearly not all price lowerings result in a price war, how can we fix our model?

2. How are initial prices for new shops picked?

3. Does current market power (in volume, price) effect network power?

4. What might be other non-price incentives?

An interesting, related problem is that of competitive facility location, if coffee shops can move their locations.

Starbucks in the news