Welcome to the course blog for MS&E 135 Networks at Stanford University. This course explores how diverse social, economic, and technological systems are built up from connections, and how the study of networks can help us understand these systems.

Enrolled undergraduate students will be writing regular posts on varied subject matters and current events related to the course. Topics include: networked markets, social networks, information networks, the aggregate behavior of crowds, information diffusion, the implications of popular concepts such as “six degrees of separation” and the “friendship paradox.”

The blog is visible to the public, however only students and course staff are able to post and comment. Students should refer to the course blog guide for more information.

How a Few Anti-Vaxxers Landed on WHO’s Top Global Threats List

The anti-vaccine or anti-vax movement is a threat to the progress that has been made against infectious diseases. The World Health Organization (WHO) lists “vaccine hesitancy” i.e. delays in vaccinations or refusals of vaccines despite availability, as one the top 10  global threats of 2019 (1). The effect of the anti-vax movement is evident in the measles resurgence in the US. Since 2010 the CDC has recorded over 2000 (2) individual cases of measles. Research has found that “a substantial proportion of the US measles cases in the era after elimination were intentionally unvaccinated. (3)”

How has the anti-vax movement become a global healthcare threat despite more than 90% of 19- to 35-month -old American children being adequately vaccinated?  When immunization rates fall, herd immunity can break down and leave susceptible individuals exposed.  Herd immunity arises when a significant portion of a population provides a measure of protection through vaccination for susceptible individuals. Herd immunity can stop the spread of infectious disease as there a few susceptible people left to infect thus providing critical protection for people who cannot be vaccinated.

Anti-vaxxers spread their ideology through social contagion and information cascades. However, the initial anti-vaxxer adopters cannot cause a complete cascade as there is a cluster of density 1-q, q being the threshold for the remaining nodes to adopt vaccine hesitancy, in the remaining network. The incomplete cascade results in several small clusters of unvaccinated individuals who when grouped together lead to the breakdown of herd immunity. “People who deliberately go unvaccinated can provide the critical mass of susceptible individuals that can help start outbreaks that vaccination would otherwise have prevented. (4)”  The independent probability p of infection is higher in these contact networks than in the overall population. The basic reproductive number R0 >1 due to the higher p. Thus while there has only been a small decrease in overall vaccination rates, these areas of vulnerability can lead to an alarming number of recorded cases of infection.


  1. https://www.livescience.com/64523-anti-vaccine-movement-top-global-threats-who.html
  2. https://www.cdc.gov/measles/cases-outbreaks.html
  3. https://jamanetwork.com/journals/jama/article-abstract/2503179
  4. https://www.washingtonpost.com/outlook/2018/11/29/why-small-groups-vaccine-refusers-can-make-large-groups-people-sick/?noredirect=on&utm_term=.0c5689a3a4ab

The Vaccine to the Anti-Vax Movement

Measles has made a comeback. According to the Washington Post, the number of measles cases in just the first three months of 2019 is the highest it’s been in 27 years.[1] So it’s no wonder that people are particularly outraged with Anti-Vaxxers as of late. The outrage directed at Anti-Vaxxers has grown so much that Facebook has vowed to eliminate anti-vaccine propaganda.[2]

I find it morbidly fascinating that the anti-vaccination movement has managed to grow so big that it’s opened the doors for deadly diseases like measles to make a comeback. Think about it in a vacuum. What could possibly turn someone anti-vaccine? It’s a ludicrous stance to have. Who would believe such a thing? Well, it’s been shown that the primary demographic of the anti-vax movement is composed of young mothers and fathers, people who received minimal or no college education, and the financially complacent.[3] It’s almost as if this anti-vaccination propaganda is a disease in of itself (a mental/moral one) that preys on overly-cautious individuals with anti-intellectual sensibilities and spreads like wild-fire throughout misinformed communities.

Given the inherently irrational nature of the anti-vaxxer movement, I don’t think it’s appropriate to just model this problem as some sort of social contagion. That would imply that the individuals who buy into the cause are choosing to do so. If I had to speculate, I’d wager that new parents are protective of their children and are much more vulnerable to fear-mongering than typical people. Uneducated people are less likely to question anti-vax propaganda when they’re exposed to it. Financially complacent parents have few real threats to protect their children from so they’re more likely to become protective over the wrong things, namely vaccines. Furthermore, if you’re not a member of one of these demographics, then you’d probably be able to better assess the issue in your head, and you’re probably more likely to consider outside sources if you aren’t sure.

This can be modeled by an adjusted version of the SIR epidemic model. We could say that the majority of Americans are susceptible to becoming anti-vaxxers, but still, only a minority of Americans have been infected by the anti-vax movement. Once an individual has been exposed to their propaganda, if they choose to move on with their life then they are removed. This makes sense because somebody who doesn’t respond to the propaganda is less likely to associate with the anti-vaxxers anyway, leading to the anti-vax communities becoming more insular overtime, perpetuating misinformation amongst each other.

Another way that the SIR model comes in handy is by modeling the types of relationships that lead to new infections. We can assign a pv,w to any pair of people v and w where higher values of pv,w correspond to a close relationship and more likely contagion, whereas lower values of pv,w correspond to a shallow relationship and less likely contagion. From this we can gather that two close friends are more likely to share the same stance on vaccination than they would if they were just shallow acquaintances. Furthermore, someone is probably less likely to become friends with an anti-vaxxer if that person is decidedly not an anti-vaxxer.

This model shows us that the way to win the battle against anti-vaxxers is by removing as many nodes from the network as possible by exposing them to the virus (the anti-vax movement) in small doses. This way, individuals are only ever infected in a controlled environment where they can fight off the virus and become immune. It’s reasonable to assume that once someone has thought rationally about the anti-vax issue and seen the evidence that vaccines do not cause autism or anything else, that person will never be able to go back on their understanding.

So, what’s the vaccine for the anti-vaccination movement? Well, we need to expose people to the anti-vaccination movement before the anti-vaxxers get to them so that we can control the narrative. We need to teach people why the anti-vaxxers are wrong about vaccines early on so that they are immune to the propaganda. We can’t ignore the anti-vaxxers. We can’t just brush them off and pretend they’re irrelevant. We need to fight the movement like we would fight a disease: with a vaccine.

[1] https://www.washingtonpost.com/us-policy/2019/03/08/is-shaping-up-be-worst-year-measles-since-cdc-data-show/?utm_term=.573fc5bed63e

[2] https://www.washingtonpost.com/business/2019/03/07/facebook-says-it-will-take-action-against-anti-vax-content-heres-how-they-plan-do-it/?utm_term=.067c19b3abec

[3] https://www.ncbi.nlm.nih.gov/pubmed/28926775

Network effects of responsibly investing

People have a psychological inclination towards risky investing: putting all their eggs in one basket as opposed to a few eggs in many baskets. This is likely why the importance of diversification is stressed so highly amongst all investors ranging from hedge fund managers to college students with newly opened Roth IRAs. According to a US News article, responsible investments, meaning diversified portfolios, make every investor better off. This can be explained in terms of network effects, where increased numbers of investors with diversified holdings has a direct benefit to all investors. This is due to the efficient distribution of resources that diversification entails, which leads to improving the economy, markets, and portfolios. In fact, when investors take big risks with their own money, that raises the entire economy’s risk level, which leads to negative externalities that poorly affect every other investor.

Herding also takes place in this context when investors succumb to their natural urges of increased risk tolerance. While this usually happens when the economy is doing well and people feel secure in their jobs and wealth, when one investor makes a risky investment, others in similar positions follow suit to keep up with their peers and in the fear of missing out on the next big investment. This is a very dangerous phenomenon because it not only leads to rapidly increased growth in the economy’s risk level, but it also leads to stretched stock valuations and growing inflation. Another example mentioned in the US News article regarding the effects of herding referred to Coinbase’s addition of 100,000+ accounts during the 2017 Thanksgiving holiday. This was largely due to the social pressure during family time with people who shared their knowledge of bitcoin with other family members who fell into a “fear of missing out” mentality and followed suit with their own investments in bitcoin. Whether or not this example of herding has a good or bad effect on society is debatable, but what we know for sure is that herding mentality in responsibly investing can very easily turn in either direction.



Article: https://money.usnews.com/investing/investing-101/articles/2018-04-30/the-right-kind-of-herding-behavior-lifts-all-portfolios


In Addition to PageRank, Passion and Authenticity are the new Factors of Google Search Ranking?

Google is best known to order the ranking of its search results using the PageRank mechanism discussed in class. Developed by one of Google founders, PageRank measures the relative importance of each web page by the number and quality of links from other pages. The idea is that more important pages are “likely to receive more links from other websites” (Google). PageRank is a compelling and powerful algorithm, but certainly over time Google has had to complicate this mechanism a lot more to better fit the ranking with increased user data and preferences, among other factors. Therefore, it would be interesting to look at just how this mechanism has been adjusted, explained in simple, real-life factors of what could possibly affect the ranking of your page.


Specifically, many data scientists (or just keen online observers) have noticed that there are certain characteristics of high-ranking pages that would just not fit with the traditional criteria of how to get a high PageRank score. For example, a high-ranking page may not have the link profile, domain authority, or social metrics to perform high on PageRank (social metrics are essentially a measure of how well-linked-to the page is). In addition, there has been a growing number of high-ranking articles that come from small, niche website, which means they are “hidden gems” that could not have been linked to by many other important pages. Along the same line, due to the growing trend of de-commercialization of the online sphere, more contents are created to be personal blogs, with the owners possibly not even being aware of how PageRank works, as many of these blogs are not created with any hashtags or keywords. Keep in mind, hashtags and keywords are in the theory one of the major factors that pages are discovered, and through which PageRank value is determined. Taking all of these seeming “deflections” from the PageRank mechanism into account, data scientist and report Rand Fishkin proposes that following changes in user preferences, Google has used more personal factors like whether web pages contain passion and authenticity to have them bypass the traditional PageRank algorithm and still show up first on the search results.


In his article, Fishkin also interestingly names a few ways in which Google may have tried to figure out the level of authenticity and passion through a codified algorithm. These factors, accoding to Fishkin, include:

  • “Design & UI Quality (possibly via quality raters or the machine learning layers on top of user data)
  • About/Contact details (looking for authentic about + contact information to confirm the site is created by a real person/team)
  • Connections to the rest of the web (social accounts, job posts, a resume, partners, clients, etc)
  • Diversity of traffic sources (authentic sites/pages get referral traffic, social traffic, clicks from emails and, yes, some search too)
  • An offline presence in the real world (how Google measures this is beyond me right now)
  • Connection to other humans (people list it in their LinkedIn profiles, in their Twitter accounts, on their business cards)”

And so on.

Through an interesting thought, this proposal is new and it is hard to find concrete evidence. However, I am inclined to think that whether intentional or unintentional, Google algorithm has incorporated more “human” factors into its ranking algorithm, rather than just a numeric calculation. The reason is that Google’s biggest asset is arguably its user database, and virtually all of Google’s algorithms (regarding search, ads, layouts, etc.) do their best to incorporate this user information. Then thinking from a societal perspective, the allure and de-individuation of the Internet and social networking sites have started to wear us out, and now we have started to seek more personalized and relatable contents. Therefore, having passion and authenticity as one of the factors influencing search rankings? Not a bad idea.






Six Degrees of Separation in Popular Culture

The “Six Degrees of Separation” (SDS) is a small-world phenomenon which illustrates that the world we live in has an incredibly well connected social network with many short paths from one person to another.  As we have learned in class, the SDS is the idea that everyone is connected by six people or fewer. The concept of SDS is incredibly prominent in pop-culture; there have various movies, television shows, songs, and, even, a video game centered around this theme.  In my blog post today, I would like to examine two different cases of SDS utilized in popular culture.

Earlier this year, Netflix released an advertisement stating that we are only six shows away from any Netflix user anywhere in the world.  This sounds spookily similar to SDS, so how did Netflix actually discover this phenomenon in its service? Netflix used the vast data and streaming patterns of its 139 million person user base, including residents of 190 countries.  A user had to either watch 70% of a movie or at least an episode of a television series to count as having watched it. After accumulating this data, Netflix discovered that “The average number of shows shared between two randomly paired Netflix active accounts is 6.2.”  Many of the common films between users were big box office titles such as Stranger Things, Bird Box, La Casa De Papel, Black Mirror, and The Christmas Chronicles; these were in the upper quartile of most commonly watched films. I believe that this finding can be perceived in two different ways.  Primarily, Netflix has provided a service that caters to many people and allows the opportunity to connect with others through shared movie choice. Secondarily, Netflix has become so dominant in the industry of online entertainment that it has herded people into watching big box office films and has hurt the popularity of smaller films and independent media.  Although I am not particularly swayed to one side, it is comforting to know that someone across the world from me is simultaneously cuddling up and enjoying a nice episode of The Office.


In 2018 Jacob Wenger, a developer who has worked at Microsoft, Firebase, and Google, created Six Degrees of Wikipedia.  It is a game on the internet that allows a user to type in any two Wikipedia article topics and it will produce the shortest path from one to another.  It uses the data from the 6 million different Wikipedia pages to sort out the path and proceeds to map out the path visually as well. According to its first blog post, it had over 500,000 searches within the first two weeks.  Wenger found that there were many, surprisingly, popular searches. The most commonly searched links were Anime to Obesity, Age of Enlightenment to Consumption of Tide Pods, and Anime to Alt-Right. The most common pages were Adolf Hitler, Donald Trump, Barack Obama, and Jesus.  Finally, there was only 1.07% of searches that could not be connected via any path. Play it here (https://www.sixdegreesofwikipedia.com/). Additionally, Facebook, LinkedIn, and Twitter have also been proven to exhibit similar signs of SDS.  


After reading these articles, I cannot help but sing in my head “It’s a small world after all.”


Learn more about Netflix SDS:

https://www.forbes.com/sites/danafeldman/2019/02/13/according-to-netflix-we-all-have-6-shows-in-common/#44691ce463d0 https://www.ibtimes.com/6-degrees-netflix-how-users-are-all-connected-2766847

Learn more about SDS Wikipedia:



Dragon Year Baby Booms : Truth or Hoax?

With each Chinese New Year celebration comes questions of relatives getting married, adopting new jobs, and having children. Since ancient times, dragons have been revered in Chinese culture. With this association of nobility and good fortune, it is a common belief within Chinese culture that children born in the Year of the Dragon are especially destined for success and power in life. In an article by economists Naci Hocan and Han Yu at Louisiana State University, the authors note that Chinese parents have strong preferences towards raising children born in the Year of Dragon versus any other year. This trend is most prevalent in countries with a strong Chinese diaspora, which include China, Hong Kong, Taiwan, and Singapore. On average, the number of Chinese births increase by 9.3 percent in the year of the Dragon, a change that is not reflected in any other years in the Chinese calendar cycle. In 2000, Hong Kong saw more than a 5% rise in the number of births with the year of the dragon.  

Information cascades have potential to occur in sequential decision-making, oftentimes originating from present decision makers observing past actions and inferring certain information from these previous decisions. When we look at the origins of why this superstition came to be, there seems to be little ancient history that substantiates this correlation that being born in the year of the dragon results in relative success. The dragon has always been regarded as a symbol of might and intelligence in Chinese culture, yet this has somehow manifested in the belief that children born that year will reflect the same qualities. Rather than past studies or data, this study has been perpetuated by an ingrained belief in tradition passed on from ancient times and word-of-mouth within the Chinese community.

Rosanna Xie in The Los Angeles’ Times notes that “although most ancient superstitions have disappeared from the conscience of second- and third-generation Asian Americans, the belief that the Year of the Dragon is lucky has persisted”. Since ancient times, it has been a trend among Chinese cultures to support this belief, even to the degree of manipulating marriage times to ensure their children are born within that time frame. While the trend is driven by the influence of deeply ingrained superstitious beliefs, even couples who do not fully buy into the mythology of the Chinese zodiac system still prefer Dragon babies. This preference of dragon babies is built off observations of other Chinese couples following the same behaviors throughout history, even adopted by those who are less superstitious. Fertility decisions are more or less influenced by the behavior of fellow peers instead, demonstrating this herding effect as more couples hear that siblings, cousins, and friends are preparing their pregnancies to align with the year of the dragon. These influences are then largely rooted in social pressure and tradition rather than proven correlation. While most younger, well-educated parents do not completely subscribe to these astrological beliefs, there is an appeal in satisfying older generations who tend to more closely follow these traditional superstitions. This superstition has even manifested commercially, with several companies within the China’s fertility industry offering discounts and specials to account for the year of the dragon, further heightening pressure on couples. An article by BBC News even claims that the lure of the dragon year has even influenced prospective parents outside of Chinese culture. Rather, their motivations are driven by the fact that given the Chinese population has been subscribing to this belief for so long, so there must be some truth to the longstanding tradition.

Ironically, we see this influx of births may have  an adverse effect on the generation born in the Year of  Dragon. The article states that a key explanation for these less than ideal outcomes is that given so many babies are being born in a given year, there is more competition for resources. Dr. Zhang Yanxia, a research fellow at the East Asian Institute in Singapore, notes that the dragon baby boom “will put a lot of pressure on hospitals, kindergartens, and schools in China”. Regardless, for parents who are able to enroll their children in the top schools, these families seem to be investing more time, money, and effort into cultivating their children’s education. In an article by The Economist, the author claims that parents of dragon children are relatively more involved in educating their offspring, even giving out less in-home responsibilities to allow their children to focus on their education. Many critics argue that the whole superstition has led to a self-fulfilling policy in that the children are not necessarily more talented or skilled, but receive more guidance, support, and confidence from their own parents. When accounting for factors of parents’ increased investment of time and energy to ensure their children live up to this superstition, this academic advantage seems to disappear.








Honey Bee Epidemics

As we learned in Chapter 21 and on Tuesday, we can drastically improve the prediction of network behavior through the friendship paradox and the structure of networks. Most network behavior regarding information, sickness, etc. can be described with the S-shaped adoption curve. By using the friendship paradox and other indications of network structure to target central nodes, we can move the adoption curve to the left when graphed as a function of time. In other words, we can detect the beginning of the spike in slope of the S-shaped curve at a significantly earlier point then if we use a random sample. In order to find nodes that are more central in the network, researchers can have groups list their friends, and find the intersection of peoples’ lists. This is useful to help predict epidemics faster.


The most common — and possibly most useful — application of this idea is to predict sicknesses, viruses, and epidemics in society. However, especially on a college campus like Stanford, there are many more entertaining applications. For instance, the idea of predicting network behavior earlier by finding more central nodes can be applied to the concept of fashion. By looking at the leaders of the fashion industry, who are constantly pioneering new looks, we can predict the popularity of items much faster. We can also apply this to drug use. By focusing our attention to central nodes in the network of drug users, we can predict new drugs and develop precautionary measures. Likewise, we can do the same thing with STDs. Of the network of people who are sexually active, we can monitor those who are the most sexually active to predict STD outbreaks earlier.


One somewhat abstract application of epidemic monitoring does not involve humans at all, actually. In recent years, we have become aware of the declining honey bee population. Many describe it as colony collapse disorder. However, if scientists we able to model the network of honey bees, which can roughly be done through general population and location metrics, they could predict the areas in which the population will suffer. In simpler terms, had/have scientists modeled honey bees and thought of their population as a network, they could target more populated regions of honey bees and monitor their health. By observing the decline in population of central nodes/groups within the honey bee network, scientists could have a better understanding of the causes, timing, and general order of honey bee decline. This could help mitigate the effects of climate change and could be applied to many different animals for the promotion of a healthy ecosystem. By doing so, scientists could bypass the ethical questions involved with the modeling of human networks and potential exploitation of information.



Honey Bee Colony Losses 2017-2018: Preliminary Results

An Internet Epidemic? – The Momo Challenge

A recent craze that’s been flooding the internet in the United States is the Momo Challenge. The Momo Challenge is a social media game where kids are being targeted with an image of a creepy chicken woman figure that will be followed with messages to do destructive things to others and themselves. It has led to many parents to be concerned as they are terrified that their children will come across this challenge on the internet and hurt themselves. The places where the challenge is supposedly targeting the most is YouTube Kids and WhatsApp. This fear by parents on their kids’ online activity has made its way to police departments and schools to put out messages warning parents of the ludicrous “suicide game” hitting the internet by storm. The fear of this challenge just keeps growing as even Kim Kardashian brought light to the subject trying to get YouTube to remove these videos from their platform. She has a following of over 129 million, and the number of parents that became aware of the subject will keep on growing as now they will now tell all their parent friends trying to protect children from the internet challenge.


As of now, there is no evidence that the Momo challenge is anything more than a hoax that is driven by fear of parents throughout the planet. Even YouTube which is considered the place where kids are to see this challenge has reported that they haven’t seen any videos coming up promoting the challenge as it would be against their policy. They took action and demonetized videos that contain the iconic Momo Challenge image depicted below.

Therefore, the Momo Challenge is just another internet scare taken way out of proportions by parents and news sources alike.


The whole virality of the Momo Challenge is a social contagion that is being spread by the decision making of parents to inform other parents of their fears which has turned into a branching process. When news sources and social media influencers inform about the topic they spread the awareness of the challenge to many people watching and their followers with high probability p. Then the conversion rate of when parents talk to their other friends about the topic is pretty high given their relationship, the fear behind the topic, and their lack of knowledge of what their kids are doing on the internet. Therefore, there doesn’t seem to be an end to the spread of this topic until the public realizes that it is all an urban legend gone wild. In the future scares like these could happen again as the internet is a vast place with people with bad intentions.





Forecasting Bitcoin Price with Network Theory

     An interesting method I observed for predicting, specifically, Bitcoin price using networks, was in the study Forecasting Bitcoin Price with Graph Chainlets; which I highly recommend checking out (the cakcora website listed at the bottom). In the study they use a heterogeneous graph model to portray a “Bitcoin graph” composed of addresses, transactions and blocks. Figure 1, taken from the study, shows an example network for 4 transactions and 13 addresses.

     Because cryptocurrency information of distribution is, by design, open to the public, anyone is able to observe all financial interactions on the network and analyze how the network evolves over time per user. The data they used comes from years 2009 to 2018, in which they parsed the Bitcoin blockchain files, and extracted blocks, transactions and addresses. Rather than observing graphs with multiple transactions or individual edges and nodes, they chose to analyze what they called “chainlets” which represent subgraphs within the main graph. This means taking one block, representing a transaction in time, with its inputs and outputs and comparing this subgraph to others.

     Ultimately, they observed different kinds and groupings of chainlets into 5 different models and paired the models to the data of how they performed from 2009 to 2018 (shown in figure 6 bellow).

    All in all, nascent (new) markets like cryptocurrency are inevitably unstable for many reasons. The main reason being, there is limited liquidity in the market, compared to a more established market like traditional economies. For example, the total value of all the money in the world is more than $90 trillion, whereas the total cryptocurrency market value is around $250 billion (a 36,000% difference). This leads to a very thin market to work with that moves quickly up and down, increasing the volatility of cryptocurrency prices. This sounds like an amazing platform to analyze and take use of in theory; though, if cryptocurrency wasn’t such a fluctuating market, it could produce more consistent results overtime. This is why graph theorists try to analyze the “extreme” cases or volatility in crashes of cryptocurrency. As of now graph theorists can make predictions for where the market will go, but with such a new developing market the risk will be inevitably high regardless of what the theorist claims.


The Friendship Paradox and Hook-up Culture

As students approach their freshman year of college, many feel nervous about the prospect of hook-up culture, which surrounds the reputation of most campuses. On most college campuses — including Stanford, the friendship paradox can be seen in numerous situations. However, in the name of comforting insecurity with data, I believe the most interesting example is shown in hook-up culture.


The Friendship Paradox can be described simply: empirical evidence suggests that on average, your friends have more friends than you do. Coined by sociologist Scott Feld in 1991, the paradox seems not to make sense, but it is a mathematical fact. The vast majority of people have a few friends, while a small number have many friends. The small group with many friends is the cause of this paradox. Because they have more friends, it becomes more likely that they will be in your group. By doing so, they have effectively raised the average number of friends your friends have. Because of this, your friends have more friends than you do — on average.


This paradox can be seen in numerous situations, especially on college campuses. It is likely true that, your friends are in more classes than you are, your friends are happier than you are (MIT Technology Review), your friends are richer than you (MIT Technology Review), that your sexual partners have had more patterns than you have had(MIT Technology Review), and that your sexual partners are more sexually active than you are — on average, of course. This likely contributes to the perceived hook-up culture of college campuses. By recognizing and understanding this paradox, we might be able to soothe the nerves of insecure students who worry about popularity, status, or even sexual activity.




The friendship paradox: Why all of your friends have more friends than you

Finding Patient Zero

Image result for patient zero network

The fear of epidemics is not a new concept. Viruses and other pathogens are constantly mutating, posing a threat to society as a whole. Recall the West Africa Ebola Outbreak in 2014: over 10,000 people died in the epidemic and fear that the virus would spread across the world was widespread. There were even isolated cases in developed nation like the United States and the UK. Nevertheless, preventing another epidemic like this from occurring is a high priority to national security and is not taken lightly.

In attempting to prevent these epidemics, learning from “patient zero” of previous occurrences can provide helpful insight. Patient zero is the term used to describe the first carrier of the disease who introduced it into the population and searching for one out of thousands can be a very difficult task.

In the early 20th century, a sanitation engineer was pretty certain he was able to track down patient zero from a Typhoid outbreak in New York. Back then, the only real solution was to extensively research patients and track the disease back in time. This method is obviously very tedious and can lead to error. However, modern network analysis has improved our capabilities to find patient zero.

Nino Antulov-Fantulin and his colleagues created a new method for finding patient zero. The model they created starts with real-world interviewing and data collection to better understand the disease. Then, a network is constructed to represent the population. Out of all of the once-infected individuals, they run a simulation using the SIR model as if that individual was patient zero. Then they compare the actual result of infected individuals to the simulation in order to solve for the probability that the individual was patient zero. This is then repeated for all potential candidates.

The SIR model, which stands for susceptible-infectious-removed, worked by classifying nodes in the network into these three categories. The model starts with some nodes in the infected state and the rest in the susceptible state. In this case, the researchers ran the model several times with one node starting in the infectious state, serving as patient zero in the simulation. Next, from the data they gather about the disease, they determine a p value- the probability that the disease will transfer to neighbors of an infectious node- and a t value- the amount of steps required for a node to move into the removed state. Next, they run simulation to simulate the epidemic. Comparing the simulations to the actual epidemic, the researchers were able to get closer to finding patient zero in their tests.

This application of network analysis will make the process of finding patient zero easier, thus helping to prevent epidemics in the future.



Network Effects in Lyft

Last week, Lyft released the filing for its IPO; the filing’s ‘risk factors’ section notably mentions the company’s dependence on network effects. Specifically, Lyft discloses that “network effects among the drivers and riders on [its] platform are important to [its] success” and furthermore that “if [it is] not able to continue developing [its] … network effects, [its] business, financial condition and results of operations could be adversely affected.” The strategy question at the center of the Lyft IPO, as well as other headlining questions about whether Lyft can sustain growth and can eventually turn a profit, focuses on these aforementioned network effects.

Network effects occur when, for some decision, one incurs an explicit benefit when aligning one’s behavior with the behavior of others. In the context of products, a product is valuable to a customer to the extent that other people are using the product as well. In other words, the addition of a new user increases the value of the product for other users. The success of many well-known tech companies, including Google and Facebook, have relied on network effects and have thus cemented it as a critical part of strategy in tech. One of the main reasons that startups place a significant emphasis on growth has to do with network effects — only with more people signing on to use the product can the product truly take off.

One of the challenges that new companies often face is getting enough interest in the initial stages. Particularly with a product that depends heavily on network effects, this is especially important. If the company cannot garner enough interest and get people to use the product, the product has little value to potential users. For Lyft, if it cannot get enough riders to use its platform, then drivers will not have incentive to use it (i.e., there are few people looking for rides); on the flip side, if Lyft cannot get enough drivers to use its platform, then riders will not have incentive to use it (i.e., there are few people providing rides). To get over the initial barrier of getting users to sign up, Lyft has used several strategies, including offering first-time users a free ride. In doing this, Lyft is not simply giving away rides for free; by gaining a potential new user, Lyft’s value has increased (i.e., more riders means more drivers will sign up, which in turn means more riders will sign up).

It is important to realize that different kinds of networks rely differently on network effects. Specifically, ride-sharing networks like Lyft’s are not as advantageous as search or social media networks. One property of networks that determines the platform’s potential for success is clustering. The more fragmented a network is (into local clusters), the more vulnerable the network is to competition. For example, Lyft, which has a highly localized network (i.e., riders and drivers only care about the drivers and riders in their city), is more vulnerable to competition than a global network like that of Google search. Another property that determines a platform’s success is multi-homing. When multi-homing (i.e., when users form ties with multiple platforms at the same time) is pervasive in a platform, it is difficult to generate a profit. For Lyft for example, many of its users also utilize Uber — and compare prices and wait times for the two applications — which means Lyft constantly has to compete with Uber for riders and drivers. These are only a few of the concerns that Lyft and its investors have had to keep in mind as the company IPOs.

Network and Cascade Effects in Battle Royale Video Games

Battle royale mode is the latest phenomenon in video game culture, with game development studios rushing to get their own battle royale out in hopes of dethroning Epic Games’ “Fortnite” in hopes of cashing in on the millions being made weekly in the genre. Electronic Arts (EA) might have just found their money-printer in Apex Legends. It is being dubbed the “Fortnite killer,” although it’s too early to tell if that is the case, but it is without a doubt its biggest competitor to date after having risen to the challenge among the dozens of other battle royale games in the market. What exactly is it that makes these games so profitable? After all, they’re free to download, so it can be difficult to imagine where any revenue could possibly come from.


The answer is microtransactions, and, get this: it’s purely aesthetics. None of what the in-game stores offer gives the gamer a competitive advantage in the arena. Rather, users are spending their money on things like skins and dance-moves for their characters in order to set themselves apart from other players. The key in all of this is establishing a player base early on. “You get that installed base in there, and you really get a network effect where revenue from the game could be exponentially bigger,” says Shawn Cruz, senior trading specialist at TD Ameritrade (1). “You get the initial adoption, and what you have to do is turn that into a stream of cash flow.” 


To establish this player base, EA paid gaming influencers like Ninja and Dr. Disrespect to be the first in playing Apex Legends live on Twitch to build hype for the new game and create a cascade effect around the world. Given how well connected they are in the gaming community and how many millions of followers some of these influencers have, their behaviors were quickly replicated by gamers all around the world, hitting 1 million total players in 8 hours, 10 million in 3 days, and 25 million in a week (2). You don’t even have to be a follower of these influencers; as long as one of your gaming friends is, they will adopt the influencer’s behavior, and you will soon follow.


As people began to notice player-counts rise, more and more followed by downloading the game. Getting that initial adoption is what has been the struggle for countless of other battle royale games. Smaller player bases for games discourage others from downloading the game, since battle royale modes usually involve over 50+ users in a lobby playing concurrently. So the more gamers that are playing a specific battle royale game, the more value that is added to the game itself, both through player experience due to having so many other players playing the game and contributing to this cultural phenomenon, and through revenue for the game studios that are able to reinvest that money into improving the game continuously and give players more unique microtransaction options.  



  1. https://www.benzinga.com/analyst-ratings/analyst-color/19/02/13211172/a-network-effect-new-model-for-video-games-relies-on-in
  2. https://www.forbes.com/sites/insertcoin/2019/03/04/apex-legends-just-hit-50-million-players-in-its-launch-month-far-outpacing-fortnite/#716f1b4f4d5c

How the “Rich People Network” Feeds the Poverty Cycle in Thailand and Asia

In the article “Rich get richer, poor get the picture,” Thai writer Paritta Wangkiat complains of the Thai poverty cycle: “the majority are being overlooked as the door of opportunity is opened only for the rich to get richer” [1]. What she is talking about, specifically, is how the state keeps rolling out policies that keep on benefitting the top 1%: “investment incentives, tax exemption and partnerships with government projects” [1]. Investment incentives, tax exemption, and government project partnerships are given only to companies owned by those who are well-connected in Thai society and have close ties to government officials. And, not surprisingly, the “well-connected” people in this network happens to be the top 1% (or, more accurately, 0.01%). Government policies are designed and lobbied for by those in the “rich people network.” Once you are in this network, you will gain the power to put yourself in places where you can make even more money, which helps you build even more links and ties within the “rich people network,” and further increasing your wealth and strength in this network, and so forth. If you are outside of this network, it is almost impossible to insert yourself into the right places and know the right people who will help you benefit from government policies and similar situations. Special policy benefits is just one demonstration of how the “rich people network” in Thailand (and many other Asian countries) perpetuates the poverty cycle. Note that I am not saying that this system is bad or good, nor am I blaming any specific government or political party (I understand that it is very difficult to put an end to such an ingrained system, and I definitely understand how Thai politicians often already have too many important things to deal with). I am merely making an observation with regards to the surprising power of the network effect, which is a main theme of this course.

Another example is in the education system. (I am sure that a similar phenomenon exists in many other countries as well, such as Japan, China, and the UK, and perhaps even the US, to a certain extent.) All (almost all) rich and successful people in Thailand received education at Chulalongkorn University, the nation’s most prestigious university located in downtown Bangkok (or, at Thammasat University, the nation’s #2, arguably). Those who go to Chulalongkorn are granted access to a powerful network of business owners, company executives, and government officials. These connections are extremely important in Thailand, and presence in this network is *required* for career success. But the issue is, most often, students who pass the rigorous entrance examinations into Chulalongkorn are those from privileged (or at least semi-privileged) backgrounds, almost always from Bangkok, which is home to only roughly 10 percent of the Thai population. This leads to a (almost complete) closure of network of students who will become powerful in the nation. Students outside of this network, such as those from the far North or Northeast of the country, due to insufficient funds to receive a competitive high school education and geographical inability to live in Bangkok or near Chulalongkorn, have little chance of joining this great network. Hence, rich people and their children remain in the powerful Chulalongkorn network, while outsiders (mostly from less wealthy families outside of Bangkok) remain excluded and unable to join this network of great life opportunities. Successful people keep coming from the same old families, and the poverty cycle is perpetuated.


[1] https://www.bangkokpost.com/opinion/opinion/1379095/rich-get-richer-poor-get-the-picture

[General reference] https://borgenproject.org/facts-about-poverty-in-asia/

[General reference] https://asia.nikkei.com/Life-Arts/Education/Breaking-the-poverty-trap-through-education

How Information Cascades Pushed Bitcoin to the Top

Information cascades- basing your decisions on the observed behaviors of other individuals while ignoring your personal information- is a very common and relevant topic when discussing the stock market. In today’s stock market, no one individual has a competitive advantage over other individuals due to all information being public and easily accessible (if you are acting on none public information it is considered insider trading and results in federal punishment). This transparency in information sharing can cause information cascades and herding, leading to irregular and irrational market valuations. Often, stocks rise well beyond their actual intrinsic value because of the herd mentality of the market. This is most prevalent when stocks rise or fall rapidly, causing investors to follow the signal from the behavior of others and make irrational decisions. This causes the price of the stock to continue to rise or fall even further, akin to a self-fulfilling prophecy. These signals are classified as “positive” or “negative” signals and have led to what is called momentum trading, where some investors attempt to exploit sudden swings in the market to their advantage[1]. One individual might not be able to alter stock prices; however, a herd of irrational investors certainly can heavily influence the price of a stock.

A clear example of the type of ‘herd’ mentality is displayed with the recent development of the cryptocurrency market. Bitcoin is a prime example of how information cascades and herding can lead to over valuation of stock prices. Bitcoin’s price increased from $900/bitcoin to over $20,000/bitcoin in under 12 months. Bitcoin was the first successful cryptocurrency[2]. With the combination of the cryptocurrency market having no concrete valuation, the positive feedback and self-fulfilling prophecy caused by investors- who had very little information on bitcoin or cryptocurrency market, to irrationally buy into the trend. Bitcoin, along with the other cryptocurrencies- the stock became very volatile due to so many investors being uninformed and solely acting on the behavior signals of others. At the beginning of 2018, there were thousands of established cryptocurrencies, many of who are arguably much better than bitcoin, but people were focusing heavily on bitcoin because of the information cascade caused from the initial momentum of the stock inflation.  The graph below depicts the linear relationship of bitcoin’s surge in price and google hits, demonstrating the ripple effect of the information cascade and the resulting price influx from the herding. An important aspect to consider when looking at the  cryptocurrencies is the “whales” that carry much weight on the stock prices. When looking at people that have this much power, it is commonly referred to as reputation-based herding, due to the skewed weight the signals of these individuals offer to the market. For instance, Mike Novogratz- one, if not the main, advocate for bitcoin- stated in an interview that he expected bitcoin to reach a price of $20,000/bitcoin when the price was at a mere $5,800/bitcoin[3]. The same day the price of bitcoin jumped to over $7,000, demonstrating how much power certain individual’s signals can have on the behaviors of other investors.

The important takeaway being that although the stock market is theoretically a competitive and efficient market due to all information being accessible by anyone, investors still make irrational decisions by basing their evidence of the behaviors of others and ignoring or not seeking their personal information.


[1] ” The Impact Of Herding And Information Cascades On The Stock Market: Networks II Course Blog For INFO 4220.” Blogs.cornell.edu. N. p., 2019. Web. 4 Mar. 2019.
[2] Arxiv.org. (2019). Herding behaviors in cryptocurrency markets. Available at: https://arxiv.org/pdf/1806.11348.pdf [Accessed 7 Mar. 2019.
[3]Bloomberg.com. (2019). Bloomberg – Are you a robot?. [online] Available at: https://www.bloomberg.com/news/articles/2017-11-21/mike-novogratz-says-bitcoin-will-end-the-year-at-10-000 [Accessed 7 Mar. 2019].

Bartending 101: Learn How to Make a Negroni

Our class thus far has been an exploration of various networks. That is, patterns of interconnections among a set of things. We answered questions such as: Which people are more likely to become friends? Who is popular? Who holds power? We intuited paths as friendship or enmity and used the graph theoretic concept of distance to arrive at the empirical consensus that social networks are indeed “small worlds”. The complexity we observed in network structures told us a story about the social structures they seek to represent. However, what happens if we abstract away from these people-oriented interpretations of graphs and consider the underlying mathematical structure? In a non-human-centric network, which properties would still hold, and which ones would become irrelevant?

For instance, consider the art of making cocktails. In his unique blog post, Tom MacWright turned the world of cocktails into a network. Doing so required him to come up with definitions for edges and vertices that make sense in this context. The vertices, obviously, would be the cocktails themselves. The formulation of edges, however, is more nuanced. In the graph below, an edge between two cocktails indicates that the two cocktails are similar. To identify which cocktails are more similar to a given cocktail than others, we need a way of calculating the distance between two cocktails. Most people conceptualize distance in a Cartesian sense – the length between two points in Euclidean space. However, distance metrics provide an interesting way of constructing graphs from a set of elements, and different forms of distance can induce different graphs that show unusual relationships. For this application, we can view each cocktail as an (unordered) set of ingredients. Then, our distance metric needs to be a way of telling how similar one set is to another.

One way of doing that is by considering the symmetric difference of the ingredient-sets for any two drinks. The symmetric difference between sets A and B is (A – B) ∪ (B – A), i.e. all the elements that are in one set but not the other. The size of this resulting set can be thought of as a measure of how different the two sets are. Using that as our distance metric gives us this graph (click on the graph to view a higher resolution rendering):


We can read the undirected graph above as: “An Americano is a Negroni with soda water instead of gin” or “A Boulevardier is a Negroni with the gin swapped for bourbon.” We can conceptualize a “cocktail edit distance” between drinks as a set distance that gives the steps required to transform one drink into another. Remember that within this framework, order does not matter.

This isn’t the only possible metric; another way of measuring the “distance” between two sets of ingredients is via a metric known as Jaccard Distance. This metric is the size of the symmetric difference, as defined above, divided by the size of the two sets’ union. This division means that the distance metric represents the difference between two ingredient lists relative to the size of the lists; unlike with the symmetric difference metric, it will not grow larger just because the sets being compared are larger. The graphs below are projections of the cocktails onto 2D planes (using classical multidimensional scaling), in accordance with the two different distance metrics; note how they vary substantially from each other.

Jaccard distance projection:


Symmetric distance projection:


Beyond these two distances, more exotic metrics could allow us to draw more accurate conclusions. For example, when using these two metrics, the similarity of an ingredient that is swapped for another is not taken into account. However, intuitively, swapping lemon juice for lime juice is a far smaller change than swapping lemon juice for absinthe. As MacWright says in the TODO section of his post, “It’d be nice for the difference algorithm to know about ingredient types”. To tackle this problem, we would first have to come up with a way to measure the similarity between individual ingredients; one possible approach is to use a word-embedding scheme, that uses a large corpus of text to assign a vector to each word such that similar/related words are clustered together. Then, we could use a metric such as the Word Mover’s Distance to calculate the “minimum transportation cost” for moving every ingredient-vector from the first cocktail to an ingredient-vector in the second cocktail. The specifics of the algorithm are beyond the scope of this post, but I found this explanation fascinating. [1]

Most networks we have analyzed in class exhibited triadic closure which was interesting insofar that it says something about how our society works. It was a socially rooted phenomenon. This observation teases out an interesting nuance that objects don’t necessarily exhibit the same properties of network relations as people. Looking at the cocktail graphs, triadic closure is uncommon. Potentially, this is because having triadic closure would indicate clusters of cocktails that are all very similar to each other, while the list of standard cocktails likely values novelty to some degree.

However, other aspects have similar significance in both social networks and the cocktail-oriented graphs here. For example, the degree of a node can still be seen as a measure of its importance or influence; cocktails with many similar variants, such as the Negroni or the Martini, tend to be ones that have an outsized impact. By paying attention to these metrics, we can apply these cocktail networks to many important real-world questions. What is an acceptable minimum viable bar setup for a home? Which drink is most important for a novice bartender to master? Are there archetypal clusters of cocktails? Or perhaps, is there some common-ancestor proto-cocktail that has evolved into today’s cocktail panoply?

Though this cocktail edit distance analysis does not take into account cocktail building technique or ingredient type (yet) for replacements, it is nonetheless an interesting intellectual exploration – one that reminds us of the versatility of the underlying mathematical structures we explore through social networks. Mathematicians have an Erdos number indicating distance from the esteemed Paul Erdos. Actors and actresses have a Bacon number reflecting their distance to Hollywood’s Kevin Bacon. Perhaps cocktails could have their own distance: the Negroni number.

Footnotes & Sources:
[1] https://towardsdatascience.com/word-distance-between-word-embeddings-cc3e9cf1d632


The Victim of My Bully is My Friend: A Social Network Perspective on Bullying

There is no denying a major problem faced in the United States is bullying. Though the mediums through which it manifests itself has transformed with the introduction of social media in recent years, the underlying fundamentals of bullying and victim behavior has remained the same.

In a study conducted at The University of Groningen, researchers found that “Bullying ties do not occur in isolated dyads, but exist in larger networks in interplay with relations of diverse kinds. These relations in networks form complex patterns where network ties are dependent on other ties and the position of children in the network” (Huitsing). Therefore, bullying isn’t just the effects of one mean kid picking on an unlucky victim, it is an effect of a larger and more complex network and the negative ties that have been formed within it. Huitsing uses social network analysis to investigate bully-victim relationships in means to potentially find a better way to solve the problem.

Through his research, there were four main findings: (1) the interplay between negative and positive networks; (2) structural characteristics of negative tie networks; (3) child characteristics that are related to children’s involvement in victim-bully relationships; (4) agreement between informants on victim-bully relationships. For the sake of brevity, I’ll focus on the first.

The interplay between negative and positive networks “helps to understand the mechanisms that underlie bullying processes and the formation of other negative ties” (Huitsing). In the two figures below, it was found that for bullying and general dislike, two bullies (i and j) that shared common victims, had a higher tendency to like and defend each other. Similarly, two victims (i and j) who were victimized by the same bullies, were likely to like and defend each other.  Victims of the same bully form positive relationships for support and affection, while the bullies of the same victim defend each other to gain power. “The results showed that positive and negative ties in networks of children are related and that ties in one network influence the realization of ties in the other. As such, knowledge about children’s positive relations (e.g., friendships, defending) contributes to our understanding of the existence and creation of bullying relations” (Huitsing).


So what could this mean moving forward? Well, according to the study, “A social network perspective with investigations into triadic structures accounts for the group processes in which bullying and other negative relations are embedded”. Understanding the underlying mechanisms for bullying may help with bullying prevention. Educating children about group dynamics can better prevent the development of these triadic closures as seen above and thus, improve social climate in schools. Furthermore, these findings can help teachers better understand the social processes behind bullying and allow them to better detect and put a stop to bullying behavior. Though this may not be the silver bullet to ending an this seemingly perpetual issue, it could help to uncover a new approach to anti-bullying programs that target the network as a whole.

[1] Huitsing, G. (2014). A social network perspective on bullying. [Groningen]: University of Groningen.

The Herding Effect In Crowdfunding Platforms

Online crowdfunding platforms such as Kickstarter.com are filled with thousands of potential projects, and it may be hard to tell which ones will actually be successful and take off. On such platforms, individual backers select projects to which they donate money. Once a project receives 100% funding, all of the backers get some reward, such as premium access to the product. There is risk when a project has not received 100% funding: under the all-or-nothing platform model, if the project does not receive all the money, the entrepreneur returns all of the money and the backers get their cash back. They have wasted time (people have to wait months of years to realize a return on their investment) and capital, and have received no product in return. And, even when a project has reached 100% funding, the product may not actually be successful and what was promised. For these aforementioned reasons, it is important to ask how backers select the projects that they fund: it is important to somehow tell whether a project is good or bad.

However, people are actually able to see the level of funding that has been pledged to a project, and for this reason, it may actually be advantageous to model this scenario as an information cascade. Per the model described in the textbook, there is a true state of the world, in which the project is either successful and the product launches, or the project is bad and a bad product. A high signal in this case would be a promising description, or great pictures on the project page. Furthermore, actually pledging money to the project would be accepting in this case. Herding occurs when a project attracts more backers and more backers subsequently want to pitch in because people before have already pledged their money. People base their decisions off of the fact that many previous individuals have backed money, and for this reason, we expect that funding picks up as more people have previously donated. This is an information cascade: based on their own private signals and the actions of previous people (accepting), people have their choices influenced by those before them. One after another, people receive their signals, decide that they like the project, and then accept and back the project. 

At the high end of the spectrum, once a project has completely fulfilled its funding goals, it will actually receive the money and people should be more incentivized to pitch in due to the herding effect: based on the fact that a lot of people have previously given money and that the project has completely fulfilled its funding, people should be more willing to pitch in at 100% funding (since we are assuming that the cash will help the entrepreneur, and that even at 99% the project will not receive the money, even at 100%). There is indeed evidence that, once a project already has reached 100% funding, people are more willing to pitch in. This is because backers know the project already has a higher chance to be successful, that many people have already joined in, and that there is significantly less risk. You already “know that you are backing a [potential] winner,” and therefore herding has kicked in. People see the decisions of earlier people, assume that they have made a good choice, and follow in. Thus, herding should increase or especially kick in when a project has achieved all of its funding.

However, this article takes a twist and suggests that another factor may also be at play. If herding is at play, we should expect to see more funding more rapidly once a project is already over 100% funding (since people want the same perks but the risk is significantly less because the product will receive the money). There should be less funding when a project is from 95 to 100% funding. This is because when a project is over 100% it is probably already a winner. The study found that, it actually took LONGER for projects to move from 100% to 105% funding than 95% to 100% funding. They attributed this to the fact that people actually want to see projects succeed, and to help entrepreneurs. Therefore, altruistic measures have outweighed herding mentality and economic incentives. Specifically, Kickstarter projects took 2.39 times longer to move from 100% to 105% funding than from 95% to 100%, surprisingly. Even though it makes more sense economically to donate to a project that has already fulfilled its funding (because other people have vetted it and have donated their money) under the herding model, it seems that individuals are often altruistic and will donate when the project has not completely attained its funding goals.



Emotional Cascades

We all know what it’s like for a person to walk into the room and light everyone up with joy. We also know the feeling of someone entering a room and bringing everyone else down with their negative attitude. This emotional trend we have all experienced actually acts in similar ways as a network cascade. However, this emotional cascade runs deeper than just a brief mirroring of emotions.

Researchers have found that the positive and negative emotions in one person can substantially affect the people that surround them, even if they are only connected indirectly. A happy, positive attitude has been found to increase a neighbor’s likelihood of being happy by 34%, a spouse’s likelihood of being happy by 8%, and a friend who lives nearby by25%. This joy of one person can even spread throughout the network to people separated from the original happy person by 3 degrees (friends of friends)! Thus, a social network cascade begins with just one happy, joyful person and has significant effects on the rest of their surrounding social network.

These emotional network cascades can occur both in person and digitally. In text messages, it has been found that if someone is in a negative or pessimistic mood and texts their romantic partner, their partner is more likely to reflect this mood. Further, studies have shown that a decrease in the positive or negative content you read on Facebook can affect your subsequent posts. Users who witnessed a reduction in positive content consequently posted more negative and pessimistic posts and less positive ones. Conversely, those who saw a decrease in negative content had more positive posts and less negative ones. This can be seen in the graph below:



These effects have also been seen on other social media sites, like Twitter. Jeff Hancock, a Stanford researcher, believes that technology and digital sources exacerbate the anger cascade. This, he claims, is likely because screens allow the transfer of a negative attitude without having to see the person you’re affecting.

So, if you’re feeling down or blue, think about Sadness from Inside Out. When she touched any of the memory balls, they all turned the color blue as she tainted each memory with sadness. In the same way, our emotions can negatively, or positively, cascade and affect those around us in the long-term. To sum it up, choose happiness, and choose kindness!



Information Cascades in the Stock Market

In the financial world, herding behavior refers to the phenomenon by which investors will base their investment decisions on the decisions of other investors. However, in general standard stock trades in the market are not the result of herding behavior, in that individuals will buy or sell a given stock given the success they anticipate of that particular stock. Where herding behavior arises in the stock market is in the uncertainties surrounding the financial market. This articles defines three main types of herding in the markets. Information-based herding, when a large group of investors, in this case, react similarly to announced information. Reputation-based herding, when investors base decisions off of the decisions of a notorious or very well-respected investor. Lastly, compensation-based herding, when a financial institution controls a large quantity of financial assets, for securing profits, sells a large amount of stock in a sector that is commonly traded in, resulting in reactions from a large number of other investors. Usually, herding behavior in the financial markets is sparked when investors take note of unusual trading activity, trade imbalance. Upon recognition of this trade imbalance, investors will come to believe that those who were part of the trade imbalance know something they do not, so they follow the unusual activity. In a scenario where the price a stock sold for is abnormally high investors will fear that there is some reason that the price of the stock will go down, resulting in the selling of those investors stocks to avoid losing money. On the other hand if an unusually large quantity of a stock is bought, investors would will buy this stock because they believe that those who initially bought the stock had some reason to believe the price would go up.

The herding behavior described in this article can be largely related to information cascades. As we have discussed information cascades occur when the information provided by previous decision-makers’ decisions take precedence over a decision makers private information. In the case of this article the investors private information is their own evaluation of a company or a stock. Take a scenario where an individual investors analysis of a company or stock leads them to believe that a current product of theirs will decrease in popularity. This would lead them to either not buy, or not sell stock in this company. An information cascade would occur when an investor sees that a previous investor has decided to sell a stock. The investor does not have access to the previous investors private information, only access to his own private information, but decides that the previous investors’ private information supported his ultimate decision of him selling the stock, and these signals to sell are more powerful than the investor’s own private information, so the investor decides to sell as well. Now, the next investors have even more signals to sell, and will also sell. This results in a cascade that ultimately leads to the value of the stock decreasing significantly. Then on the contrary, if the initial investors had decided to buy the value of the stock would then increase drastically. This shows us that large changes in stock prices can result from a small amount of what could potentially be non-credible or inaccurate initial-decision making analysis. In the case of the stock market these cascades occur rapidly, which makes them even more likely to occur. This stems mainly from investors lack of time to do research or analysis allowing them to gain more private information. This is why the tendencies of information to occur in stock markets is the basis of stock promoters. A company does not need to convince a ton of investors that their companies stock will be successful, but rather a small group of investors so as to initiate the informations cascade.

What makes the stock markets so interesting is herding behavior’s prevalence to the markets, which makes the possibility for information cascades. In the stock market the value of a stock is based largely on investors willingness to either buy or sell a stock, at least in the short term. If a mass amount of people decide to sell a stock the value of that stock will go down drastically. The investors private information plays a partial role in the price of a stock, but this shows that news and information is power in the life of a trader and that your reputation all account for how much bearing you as an individual take on the stock market.