Wednesday, February 13, 2013

Ability or Social Institutions: Which Should be the Default Assumption?

The Economist has been publishing some very good posts on social mobility over the past week or so. One of them, however, Low mobility associated with inherited ability is no social tragedy, is problematic.

The author makes a claim about underlying mobility rates that I find plausible, though I have my doubts and it has at least hints of the Marxist notion of class which I've never completely agreed with:

If these estimates of social mobility were anywhere near correct as indicating true underlying rates of social mobility, then we would not find that the aristocrats of 1700 in Sweden are still overrepresented in all elite occupations of Sweden. Further, the more equal is income in a society, the less signal will income give of the true social status of families. In a society such as Sweden, where the difference in income between bus drivers and philosophy professors is modest, income tells us little about the social status of families. It is contaminated much more by random noise. Thus it will appear if we measure social status just by income that mobility is much greater in Sweden than in the USA, because in the USA income is a much better indicator of the true overall status of families.
 Then he makes a claim that I don't feel follows:

Many commentators automatically assume that low intergenerational mobility rates represent a social tragedy. I do not understand this reflexive wailing and beating of breasts in response to the finding of slow mobility rates. The fact that the social competence of children is highly predictable once we know the status of their parents, grandparents and great-grandparents is not a threat to the American Way of Life and the ideals of the open society.
The children of earlier elites will not succeed because they are born with a silver spoon in their mouth, and an automatic ticket to the Ivy League. They will succeed because they have inherited the talent, energy, drive, and resilience to overcome the many obstacles they will face in life. Life is still a struggle for all who hope to have economic and social success. It is just that we can predict who will be likely to possess the necessary characteristics from their ancestry.
There are several separate streams of evidence that make me sceptical of this claim. First of all, I've been reading both Sorkin's Too Big to Fail and The Romance of the Three Kingdoms. Something remarkable in each book is how many of the major players in each book place a high value on loyalty. While the characters in Romance at least have the excuse of Confucian morality as well as being only semi-historical the major players in Too Big To Fail lack these excuses, they're supposedly from a meritocratic society.

Now, while loyalty/betrayal is supposedly one of the moral foundations, it isn't one that has an equally strong influence on individuals. Assuming that ability and the strength of one's moral/betrayal foundation are distributed heterogeneously makes the idea of society sorting based on ability problematic. While individuals with a weak sense of loyalty/betrayal are likely to promote and select based on merit, without regards to variation in loyalty/betrayal, those with a strong sense of loyalty are likely to take this into account. Over time this will weaken an organization as those with strong task specific ability but with a weak loyalty orientation are selected out, limiting the talent pool to only those with a strong loyalty/betrayal foundation.

On a societal level, in the long term this will select for families which display strong characteristics of loyalty/betrayal. They will reinforce this through reciprocity that will help their scions to succeed in life, despite having being relatively weak in terms of ability. Individuals of exceptional ability may spring on the scene, but unless they are willing to engage in the reciprocal behavior, undermining their independence in the process, they will be selected out of the elite in a few generations.

A second problem is that social selections proceeds vastly more slowly than does genetic variation. An interesting study on this looked at variation in folktales across ethno-linguistic boundaries compared to genetic variation. While there is little recognizable variation in genetics across European subgroups substantial variation is shown in folk tales based on these boundaries. It takes a very large distance to lead to the same level of variation that crossing an ethno-linguistic does.

While something of a leap, it is not a large one to wonder if something very similar occurs with social networks. Anyone that has lived in a small town can appreciate this effect, it is much easier to find a job and do business with people that one has grown up with than it is to do the same when one has just moved to a new area. People from a certain social stratum are much more likely to have peers that can connect them with good jobs, give them access to activities that will get them into good schools, and once in business, have access to networks that can get them face to face with the folks they need to talk to, rather than doing long hours of grunt work to achieve the same thing.

Given the persistence of social networks and the difficulty of breaking into them compared to the high level of variation in genetic traits, the more reasonable null hypothesis is that observed inter-generational social mobility is low do to social institutions rather than individual level differences. It is possible that it is in fact a combination of nature and nurture, rather than institutional differences that are at work but this runs contrary to what we know about the strength of social networks, the observed variability in the weak measures of ability we have, as well as human biases.*

I'd also note that all of these societies mention show a high degree of continuity. As long as the social fabric remains intact elites have generally been able to maintain their social position. However, if the social fabric is deeply upset, like in some of history's larger invasions (you don't see a lot of surnames from Latin classical works that still wield large fortunes today), then this persistence seems to dissolve. It also needs to be noted the product of the state, persistence of elites occurs across regime changes as we see in the French Revolution and in changes in Chinese dynasties. Disrupting elites requires disrupting them in their social base of power to destroy the bonds of loyalty between these families, otherwise their elite position can be maintained. It does seem that their social position can be somewhat eroded through egalitarian policies that are meant to erode barriers to acquiring skills and security but bonds of loyalty and tradition mean that only the most exceptional individuals can compete against those with social advantages.

I also need to add that despite popular opinions about modern meritocracy, there is no particular evidence that exceptional ability has ever been blocked by social institutions. The Ming Emperor was a peasant, the father of the Emperor Justinian a mercenary. Looking at economic history, many powerful merchants in all societies have risen from obscure origins. The idea that our society is any better at selecting for the best and brightest has little support. We certainly reward them better, but there is little reason to think that this leads to people of more ability rising to the top; rather it seems to increase the rewards to incumbent groups. Where growth comes from isn't from rewarding exceptional individuals or society's elite, these people will do well regardless. What meritocracy can achieve is a better match between aptitude and reward throughout the income scale. This is where we are doing an increasingly poor job and part of why we're in danger of declining. The challenge isn't to reward and attract the best and brightest, it's to make sure we reward the 10th percentile of aptitude at a proportionate level to the 50th percentile of aptitude to the 90th percentile of aptitude. Every society will award the top fraction of a percent as they deserve; the question is how much of the proper reward accrues to the lower tiers of aptitude and how much is captured by the families that historically occupy the top rungs of society due to social institutions rather than native ability.

*A large bias, particularly strong in the US, is that we tend to prefer narratives based on individual effort to social determinants. It makes a better story. We tend to attribute things to intentional causes rather than diffuse unintended consequences. We also have a bias towards believing in a just world, people like the idea that good gets rewarded and evil punished and we tend to rationalize away evidence to the contrary. This plays a powerful influence on which stories gain traction and we should always be extra sceptical of a good story that confirms the status quo and exalts the powerful. These stories aren't necessarily false but deserve an extra dose of scepticism when we encounter them; especially since they are apt to ring true to our gut.


  1. People from a certain social stratum are much more likely to have peers that can connect them with good jobs

    There are two, rather different, ways that personal connections can function. One is actually meritocratic, the other is strictly about social networks and not meritocratic at all.

    Taking the second one first, it works as you say: it's all about who you know, not what you know. group loyalty is more important than anything else.

    But then there is the other sort. Here, connections can get you information about jobs that are available. But what is more important, your connections can provide recommendations because they know the quality of your work. That is, I will recommend you (probably informally) to my boss, or even to someone else (even in another company) that I know who is looking to hire, because I can say "This guy knows his stuff and can do the job for you." If asked, I can equally say "You might want to think carefully about hiring this guy, because he knows less than his resume would suggest" or "He knows his stuff, but it is going to be a challenge getting him to work well with others."

    I bring up the second kind of network effect because O know it first hand. That is how things frequently work when hiring technical people in IT. (I can't speak as certainly for how hiring managers here works.) It is how I got several jobs over the years; in fact, pretty much every job after my first. It is also something that I have done for others on numerous occasions. (We are forever saying to people "It's a very small world we all work in. Everybody at least knows someone who knows you -- and your reputation will follow you your whole career.")

    Why does it work? Because technical people want to work with others who are competent. Perhaps more to the point, they do not want to try to work with others who are not. If only because working with someone who doesn't know what they are doing means that everyone else will have to work harder to carry the dead weight. Also, if I am recommending someone, and they do not perform as described, that reflects back on my reputation -- and, as noted, that will make a difference in my career prospects going forward.

    1. I'd actually argue the second type of network effect, where incompetent people are given positions, is actually very rare. Even in the bad old days both business and governments did little more than create sinecures for these folks and kept them away from decision making. Even in monarchies, more often than not power simply flowed away from the hereditary ruler and towards other offices.

      As for more meritocratic selection, this is what tends to make it so difficult to break the powers of elites. Networking tends to be very efficient at transmitting quality signals. Where it is being used in a more technical positions this can be a channel upwards for people with less prestigious backgrounds to be noticed and moved up.

      Its effects become more mixed with more difficult to assess positions, like management. It's much harder to judge quality here and networking can play a much more powerful role in restricting who gets recognized and who gets a chance. Since little can be communicated on a resume, or by other criteria that would be relatively equal among candidates, there is little to judge on except through networking and reputation. Combine this with strong beliefs in the value of loyalty that many upper level decision makers possess, and there becomes real possibilities for group think, penalties to questioning conventional wisdom, as well as a tendency to packing organizations with people that have elite backgrounds. This doesn't close off possibilities for those of truly exceptional ability but it does create a bias towards those with elite backgrounds and makes it less likely that the top 5% or so really reflects the top 5% or so in ability, though many of these individuals will be in that top 5%. This isn't to say that children of elites in these positions would be in the bottom half of the ability curve, but somebody in the 70th percentile or so might be displacing someone in the top 2% or so by virtue of connections.

      Also, due to network effects many businesses are advantaged more by tapping into their employee's networks than they would be tapping to raw talent, further complicating the picture. Given how social structures work the value to firms is likely maximized by optimizing their use of networks. This does mean that rational maximizing behavior on the part of individuals and firms is likely to diverge from a socially optimal use of talent, however.

    2. Its effects become more mixed with more difficult to assess positions, like management. It's much harder to judge quality here...

      What you are saying (correctly, I think) is that we have very little understanding of what characteristics contribute to being an effective manager. At least, if there is anyone who has a good understanding, it hasn't spread across either managements in general nor across those which do academic management training. Obviously, if someone does work it out, they will have an enormous competitive advantage -- if only because current mangement selections are so wildly variable in the quality that they select.

  2. The author in the economist probably hasn't heard of Gaulton and the regression to the mean...The Bernoulli family is the exception rather than the rule TE wants to suggest.

    Also, I am sick and tired of discussing income as a proxy for economic status. Wealth is seldom discussed and brought into equation. At most home ownership is used as such a proxy - better than nothing, I do that myself - but it is not enough.

    1. Excellent point. Wealth and income are different issues with different related problems and solutions to those problems.

      I do think that part of the issue is that we have much better data on income than we do for wealth; though it's possible that someone is trying to deal with measurement issues. I'm not very familiar with anything but income studies, however, so can't really say.