Social Capital and Current Scoring Methods

Social Capital is an integral part of economics, but is poorly understood. It holds that social relationships have productive and economic value - something we have all known since childhood when our parents fed, clothed and taught us. However, calculating that value is exceptionally difficult. I will outline some of the current scoring systems and methods which I think are very cool and then I will point out some things we can improve as well as some new uses of social capital information.

Current systems, the Big 3:
There are 3 big current systems for social capital scoring: <a href="http://kred.com/rules">Kred, <a href="www.klout.com">Klout and <a href="http://www.peerindex.com/">PeerIndex. A video explaining how my favorite, the Kred score, generally works:

[youtube http://www.youtube.com/watch?v=AKTEXlGuxgM&w=560&h=315]

Kred is known for being the most transparent in its calculations of the score, Klout is know for having the most sophisticated scoring system, and PeerIndex is known for allowing social consumption. That means lots of cool free stuff to those with high scores.

PeerIndex is cool for their social consumption, but there index is not as good as Klout's or Kred's and therefore any consumption based on it is less than optimal. Furthermore their calculations are hidden which can lead to moral hazard as I will explain later.

I understand the motive for hiding scoring methods: You may get people who "cheat" to up their score. An accurate measurement is not bothered by those who only wish to up their score because the only manner of doing so would be to increase the real value of the thing measured. This is where these companies fall short. They use approximate measures open to cheating. Klout combats the cheaters by hiding the measurement system, but this is modestly effective at best, and produces moral hazard.

I think Kred has it best in theory. Kred's calculation is open for all to see, however this opens it up to the cheating problem because their measurements are not optimized. On the other hand they can modify rules and see how the cheaters react and so can improve the system over time, yet it has far from been perfected. Another point is that Kred provides more than just an influence score, and they consider more than just volume of interactions in calculating their scores. Finally, Kred also prevents moral hazard because there are no people behind the scenes who know how to work they system - everyone knows how to work the system! One problem with Kred is that nearly all of its source data is Twitter-harvested. It considers other sources but not as effectively.

Social Capital needs to not only measure content exposure but also whether the reaction to such content is positive or negative and to what degree. The good news is that they are working on this, but there is a long way to go. I know Kred and Klout are both trying to find ways of gauging whether reactions are positive or negative by textual and contextual analysis. In other words, as an example of textual analysis, if a reply says "good" or "love" it is textually significantly more likely to be a positive reaction then if the reply or comment says "hate" or "stupid." As an example of contextual analysis, a like on facebook or a retweet on twitter is usually good while a comment or reply may or may not be, and the negative implication of a thumbs down on Youtube or reddit has to more than outweigh the value-neutral, but influentially positive, significance of a mere view.

Some organizations correctly respond that they are not measuring social capital, but are only measuring "influence." Influence, while connoting positive social capital, is really only value-neutral. You can have positive or negative influence. This is why the organizational response is technically correct but not optimal. This is similar to when I talk about information exposure in my article <a href="http://thegenerallifeblog.blogspot.com/2013/01/trading-news-and-bitcoin.html">Trading the News and Bitcoin. Information being spread, which is measured as "influence," will only accelerate the previous trend, regardless of whether such trend was up or down in value.

So how do we correctly measure social capital? Not only by number of interactions a person conducts, but also by how those reactions are seen as good or bad. For a single interaction:

s = (m)(r)

Little "s" is the social capital change from a particular interaction. Social capital change is equal to the social capital of a person encountered adjusted for the response. The response is the degree of how much the other person likes you as a result of the interaction, to put it roughly. If they like you enough to give you a $10,000 car, for no other reason than just because they like you, then your social capital for that interaction would be $10,000. Or would it? Your social capital "with respect to" that individual (read "partial derivative") would be $10,000, but there would also be a market-wide social effect. To the degree that the market is aware of the fact that the interaction took place, the market would perceive that you are worth $10,000 to person A and consequently your value would increase in the eyes of the whole market as the value you hold to person A adjusted for the value person A holds to the market.

It should be noted that while we assume full and free flow of information in theory, in practice this rarely holds. So someone's social capital is not harmed by an affair or bastard child if he covers it up in secrecy, but in theory it would. It should also be noted that if person A and B are in the same market, which they must be to interact, then person A's perception of person B will effect little "m," the market social capital value of B. In order to make this position tenable, we must assume person A's perception of person B to be nearly neutral, or close to 0. This is equivalent to assuming that person A will be acting rationally and objectively, ignoring personal feeling about person B. This is an awesome insight to the way that free markets work! We must be cold and calculating, disregarding social or personal affections, in order to produce an economically efficient result. Is this good? Most would say, "No," except those Wall Street traders and professional entrepreneurs who have known this secret for years.

One implication of the cold rationalization of the individual as necessary for an optimal transaction means that markets in general should be cold and rationalizing. Ironically this means that social capital should have no noticeable effect to begin with. However this is noticeably false.

According to the influence-only metric Saddam Hussein had much "influence" because he was frequently discussed in social media and the news. However when he talked most people usually disagreed and wanted the opposite of what he wanted. This is something that current measurements of Social Capital are missing. His "m" would therefore be <0, rendering an accurate calculation of his social capital negative despite an arguably high influence or exposure. This also shows another interesting dynamic: If you are ridiculed by Saddam, as evidenced by a r<0, the market would react by increasing your social capital because the market holds m<0 with respect to Saddam! Now we are starting to see the interaction of culture, economics and even politics! Ya baby! Ok let me calm down.

Typing in a blog makes complex math hard to express. It would be more accurate to say that social capital is not the result of one interaction, rather, it is the collection or stock of the social capitals yielded by all interactions or encounters over time. For example if I talk to Joe, we have had an interaction or encounter. He may leave liking me more or less or being relatively neutral, but he will be aware of whatever information I disclosed in our conversation. That would be one interaction or encounter and shown by the lower case example above. Over the course of all transactions though, we will use capitalized letters to represent collections of values normally shown by a sigma equation. I will also try to express the sigma equation though it will look ugly in text:

S = MR
S = (Sigma,From t=1 to t=L,(m(r)))

Where t is the "encounter number" arranged by the order of encounters over time. L is the number of the last encounter. Little "m" is the social capital of the person being encountered and little "r" is the response value. Notice that S, social capital, is a cumulative quantity not a rate nor an average quantity. This should be expected because positive relationships early on in life will have lasting effects.

We can calculate this social capital by assessing change in perception or by noting price differentials. Ideally these two methods should verify each other because either method is flawed on its own. Assessing perception would be done by saying, "How much do you like Joe from 1-10?" Then an interaction would take place and the same question asked. The change in perception would be equal to little "r." Price differentials would be noted by having Joe go shop for cars. First he would go to 30 car dealers who he did not know and obtain the normal price of a car. Then he would go to 30 car dealers who he had a personal relationship with. Ideally those personal relationships would all be of the same magnitude and same direction (ie 3 "1-10" scale points positive or negative from a neutral 5). The difference in price would be presumably due to social capital. If the price from his connections were lower he would be said to have positive social capital and if they were higher he would be said to have negative social capital. In fact the social capital could be precisely calculated if all 30 networked dealers had the same magnitude of relationship, but not otherwise, because we would need a statistically normal sample at every relationship level in order to accurately calculate the degree of social capital in addition to the direction. Direction being "positive social capital" and degree being "$100 worth of social capital."

Social capital also should not be confined to relationships. I would call what we have discussed previously, "Relational Capital" or "Network Capital." Perhaps the second term would work better so as not to confuse the variables in my equations below. In economic theory a person's skill-set contributes to their social capital as well. This is called their, "Human Capital." For example, education increases human capital in theory. Personal fitness would play a role as well. Not because these things make people like you more, although that may be a extra effect, rather, because these things make you as an input to a production function more effective in generating outputs. A strong, healthy farmer can work the land better than a weak, diseased farmer.

In conclusion my final model would look like this:

S = N + H + J
Where N = MR, H = G + g(T,D)
J = Interaction term
g = Genetically predisposed growth function
G = Genetically predisposed human capital at birth
T = Time
D = Efforts to train or develop human capital

Social Capital is equal to the sum of Relational Capital and Human Capital, adjusted for the interaction they certainly have. Network Capital is the collection of the individual interactions people have, adjusted for the reaction value, also called the response value or result value. Human Capital is a combination of genetics and training and development over time.

In conclusion much information is required to generate an accurate score. Then the question is raised, do we really want to have that much known about us? As an economist I would respond, "Of course you do, as long as you get enough in return for the information." The question turns to you: What do you want to get out of our Social Capital Scoring future?