China: Unintended Consequences

From the Freakonomics website, an explanation for the high savings rate in China:

Some say that a major cause of the U.S. housing bubble was a surge in savings overseas, particularly in China, where the personal savings rate soared to 30 percent of disposable income. (In the U.S., meanwhile, we were saving next to nothing). Just why the Chinese were saving so much has been a puzzle to many economists. Now Shang-Jin Wei and Xiaobo Zhang think they’ve come up with an explanation. It turns out that China’s “one child” policy, which created a huge surplus of men in the country, has driven up the cost of getting married, as more and more men compete for fewer and fewer women. To keep up, families with sons have been holding off on spending to save up wealth that boosts their children’s marriage prospects. In their paper, Wei and Zhang argue that Chinese marriage-price inflation could account for as much as half of the increase in the country’s household savings since 1990.

Life is full of surprises. The problem with rational problem solving is that in the real world we usually don't have an accurate model of how things work so we constantly get "unintended consequences". Who would have thought that a one-child policy in China might lead to a housing bubble and financial crash in the US?

The lesson to be learned: be cautious in proposing "fixes" to problems. Without sufficient prior experience you are probably setting yourself up for failure. Even with prior experience, unless it is relevant and exhaustive, you still might be setting yourself up for failure.

The problem with modeling: a model is a simplification of the real world down to "essentials" which allow you to experiment to "understand" the real world. But how do you know what to leave out of the model and to what extent to "simplify"?

If you don't fully understand the phenomena, a model can't claim to be validated. But if you fully understand the real world, then the model is usually not needed. Models are a bit like the legwork in science. You explore for "facts" and develop a theory, the model is the crystalization of your theory. But like all scientific theories, experience can falsify the theory/model instantly. So the theory/model is only as good as the long series of testing you've put it through and how well it fits into a web of other theories/models that support it.

I still carry with me the interesting model that a chicken has of the world. This is from Bertrand Russell's book The Problems of Philsophy in the chapter "On Induction". He points out that domestic chickens build a model of the world where when the sun comes up then a man comes out and feeds them. This model works wonderfully well until one day the sun comes up and the man comes out and wrings the chicken's neck. The model is wonderfully accurate until it completely fails. The underlying problem is induction. How do you find "regularities" or "laws" or "model rules" on which to build your knowledge. The fact that up until now something has always been true is not logically sufficient to make it into a scientific law (or a sound model). Modelers and amateur "scientists" beware!