I have been loosely following Jim Manzi’s critique of economics. He has been addressing head on the physics envy of the economics profession. His argument to me is pretty simple: economists may indeed have insight to offer into various discussions but those insights should by regarded for the most part as the inferences of specialists as opposed to the strict application of principles that a technician would use. A lot of the folks don’t seem to understand the distinction he is making or at least they don’t seem to want to understand the distinction. This is understandable given that we are undergoing what I perceive to be a slow paradigm shift.
Perhaps others will source it elsewhere, but Newton seems as good of a place as any. His proposal that all events in the natural universe can be deduced from universal laws put us down a path of privileging deductive knowledge. Personally, I have no issue privileging such knowledge, and I don’t believe Manzi does either. The problem lies when we make circumstances into laws or as Mark Twain might say, “The problem is what we think is true but ain’t.” Take the example of the woman who has eight boys in a row. The prediction that her ninth child will be boy is reasonable enough given the data set. If we were dogmatic over the matter, we would describe invisible forces at work that caused her to produce a girl were a girl to be produced. This sounds ridiculous to us because we believe that the sex of a child is biologically determined by the man’s contribution.
I think Microsoft Office and the statistics establishment bear a lot of the blame here. Some statistical claims are better than others. Insurance has depended on the claim that the variation in the incidence of a set compared to the group incidence approaches zero as the size of the set increases. An example of this would be life insurance on 35-year-old men. A set of 10 such men may see none, one, two, or even three of them pass away. A set of 1000 men will see pretty close to the group norm. Some might argue this example is trivial, but a billion dollar insurance industry operates on this principle. Other statistical claims are not as good. Take the welfare of children without a father in the home. One could simply compare incident rates among children in one group compared to the other. Many argue that is a poor measure and suggest controlling for race, age, income, and other factors. My purpose here is not to adjudicate that dispute. What I will claim are the influences of household structure, income, race, age, and what not on children are using the veneer of statistics to advance ideological agendas. Whether or not to control for the age of the parents is not a statistical choice but does affect the outcome of the statistical analysis. The analysts are in fact deciding what are independent and dependent factors. I can hear the yeah buts from here, so let’s take income. For example the argument for controlling for income becomes nonsense if we agree – I make no claim that we should agree – that divorce causes poverty. Or to put it a different way, are we not speculating about a world that isn’t if we attempt to construct our analysis around a sample of income equivalent mothers? What is kind of funny about the whole thing is that the argument sheets switch sides when we talk about the income gap between men and women. Those that believe we need to put controls on the data all of the sudden question the legitimacy of putting any controls on the data and vice versa. Such is not to claim that the sociological community has no insights to offer.
Manzi offer the following challenge to economists:
My challenge would be simple: please list 14 useful, non-obvious predictive rules that economics provides that have survived rigorous, replicated falsification trials.
I thought I would take a crack at this given that my faith in the economics establishment is pretty low. Call it a test of intellectual honesty.
1. Where the added cost of one more item is trivial (a mature market), price is demand driven.
2. The value of a network grows exponentially as nodes are added until saturation. An example would be the Internet.
3. Bureaucracies are more efficient than piece work on an aggregate basis. (The exceptions tend to take longer, but the ordinary are significantly quicker.)
4. The use of slaves causes economic inefficiency and was the reason the South lost but paying poverty wages is good for growing economic wealth.
Okay, I was kidding about the last one. I’m curious if our readers can add some to the list. And the slow paradigm shift I’m seeing is growth for the appreciation of inferential knowledge and an abandonment of deductive models that don’t actually model the world as we see it.
Ad-Free & Subscriber Log-In