21 Sep 2011

Golden culture

Some people dislike economics for its central emphasis on individual behavior motivated by rational self-interest. They see these words and assume that economists believe in some form of every-man-for-himself cold calculation.

Economists have not helped with this impression: many propose theories of behavior based on "homo economicus" models in which individuals are only concerned with improving their own situation -- even if others should suffer.

These models are used for two reasons: first, they fit into simple mathematical models that can be easily manipulated to produce "results." Second, they fit into a shallow reading of Adam Smith's notion of an Invisible Hand promoting the Wealth of Nations (1776), i.e., people looking out for themselves while trucking and bartering will increase the wealth of everyone by providing opportunities for trade.

Non-economists have had a hard time arguing with this asocial "methodological individualism" because the math seems to be logical and unassailable (the complexity of the math doesn't help), and it's hard to defeat "U_i = f(x_1, x_2)" with "that doesn't feel right."

Fortunately, other economists have been exploring the realities of these theories. "Good" economists know that individuals include the happiness of others within their version of "self interest." These economists have used experimental economics to show that predictions in the simple math models are just not right. (I think that my simple model makes WAY more sense.)

All of this text is a lead up to a simple short story: Last night [16 Sep], I left my gold ring on the gym key that I left hanging in the locker, right next to the men's room exit. (I took off the ring to work out.)

My mother gave me that ring in 1987, and it's worth a lot to me. It's also worth a lot for its gold content -- at least $500+ at any gold dealer.

So there were three possible outcomes for when I went to the gym this morning:
  1. The ring was still hanging (unlikely, given the dozens of guys who would walk by it after I left).
  2. Someone took the ring, to sell it.
  3. Someone handed it into the front desk.
Now ask yourself what you would do if you found that ring -- and also ask yourself WHY you would do that. Perhaps you would take the ring and sell it, because $500 is handy and finder's keepers. Or perhaps you would ask yourself how you would feel if you lost a ring or how you would act if you were watching yourself as an outsider -- thought experiments that Adam Smith explored in his 1759 book, The Theory of Moral Sentiments (see this and this post).

In fact, the relevant question is not what you or I would do, but what would the average guy do who was leaving the locker room. It's clear that some would take the ring home, but also that others -- probably a majority -- would turn the ring in. The ratio between selfish guys and social guys was not really relevant last night -- the first guy to see the ring was the guy that mattered -- but it IS relevant when it comes to talking about repeated social interaction.

Put differently, we treat strangers according to the average mix of selfish/social values in the people around us. If most people are selfish (for cultural, economic or political reasons), then we tend to expect selfish and act selfish. If most are social -- treating each other as extensions of themselves -- then we tend to assume that we should also act that way.

The implications of these mixes are obvious, and have been deeply explored in experimental labs: a group of selfish people that fights over the spoils will end up poorer than a group of social people who cooperate to be fair to each other. (This discussion underlies Part 2 of my book.)

Those actions determine the Wealth of Nations in the long run; in the short run, they determine how safe, happy and generous we feel.

Bottom Line: I feel good this morning because someone turned in my ring :)


Eric said...

Linear models of 'homo economicus' were the best that could be done before computers were common. The math of linear models can be done with a pencil and, on average, these models work when the predicted system behaves linearly.

The recessions and bubbles are non linear so, often, the linear models, by their construction, can't predict non-linear events. In mathspeak, the partial derivatives of many of the variables that are important to economic behavior are set to 0 in linear models to make the math easier for pencil pushers.

Now, with well thought out computer programs, we can incorporate the dependence of economic measures on these variables that were ignored in previous predictions.

The prediction may be linear over many economic conditions and dramatically non-linear under certain conditions.

My hope is that well thought out non-linear prediction models will start to get both the linear predictions and the non-linear predictions right.

For what I do--learning and memory--non-linear models (often agent based) give linear predictions under many conditions and very non-linear predictions every now and then. The underlying model is completely rational but large and non linear.

I would like to see some predictions of carefully done agent based economic models.

Jay said...

Excellent post and excellent comment by Eric.

I second Eric's interest in agent based economic models. In the engineering world - where I come from - agent based simulations are providing more realistic representations of observed behavior, in such disciplines as traffic modelling, than the traditional aggregate models.

Each economic actor has a unique utility function, which is non-linear, and varies continuously. The term, "sensitive dependence on initial conditions", originally coined to describe weather phenomena seems perfectly suited to economics. The math will be daunting. Simulations will be the only way to proceed. These will not yield exact answers, they will provide probabilistic ranges of behaiviors.

For additional background reading - though the books and authors mentioned are excellent - I recommend starting with "The Selfish Gene" by Richard Dawkins. "The Moral Animal" by Robert Wright is also excellent.

Eric said...

@ Jay,


Both "The Selfish Gene" and "The Moral Animal" have been supplanted by more recent research.

The more recent research has shown that interactions beyond the gene level--in groups of genes, organs, entire organisms, and groups of organisms--contribute strongly to emergent behaviors of the group. Since January, I have given three courses on the interaction of morality and modern biology.

The paradigm which is becoming useful (really an extended agent based model) is that agents, each operating under fairly simple local rules, result in predictable, testable, complex emergent behavior.

Emergent behavior might be a great topic for this blog. It would take a lot of posts, though.

David Zetland said...

@Eric and Jay -- I agree that agent models can produce linear and non-linear results, but they need to be following optimizing on a single variable (like a traffic problem). Multiple equilibria models fail to predict behavior b/c agent interactions can be pushed by "random"behavior that shows up at a given moment. That's why experimental economics is interesting: it works with the black box of individual interactions, but without needing to specify the dynamics taking place in the box.

RM said...

One of your better posts (and you have many good ones). I think your math model does make way more sense. It makes room for the fact that people engage in acts of charity out of their own self-interest. For instance giving a couple bucks to a guy asking for money could happen because it makes me feel better about myself as a person, makes me feel morally superior to those who don't give, makes me feel like I helped someone, or because I thought it was funny that he was honest and told me he needed the money to buy a 40oz bottle of Steel Reserve. A person who gives always gets something out of it, so really it is an exchange.

However, I'm not sure that I agree with your idea that our behavior is based on what we think others will do. Although I'm sure there are people who behave this way, I don't tailor my behavior to my expectations of others around me, nor have I taught my sons to do that. I think it is important to do things based on what you would like other people to do for you on the theory that you can't blame other people for not treating you right if you don't treat them well. In other words, I base my actions on what I would like people to do regardless of what I think they will actually do. I say this as someone who lives in a neighborhood that many people would describe as a bit rough.

Interestingly, I find it easier to be nice to total strangers than to my own kith and kin, who are most likely to bear the brunt of my moral failings. Not sure why that is.

Glad you got your ring back. I know for myself what it's like to have something of sentimental value stolen.

David Zetland said...

@RM -- glad you like. FYI, Smith's "other" in ToMS was a version of self, i.e., behave as if you were watching yourself.

I also act according to my best ideals, but I try to take other's thoughts (and responses) into consideration.

Eric said...

The short version of my comment is that if you do not specify the dynamics inside the box and you assume linear dynamics, you will never predict non linear events like crashes. You will always be caught by surprise and will lose a lot of money.

Also 'random' events are part of any decent agent based model.

Post a Comment

Note: only a member of this blog may post a comment.