The cloud of unknowing

My father was a political science major in college. The name of that major amuses me because politics is obviously not science. In science we try to isolate variables to understand complex systems. In politics and policy there is almost no way to do this. There are too many variables to be able to isolate one. If we raise taxes will the economy grow or shrink? No way to predict because the economy is so complex it is impossible to isolate the effect of one variable. Thus Reagan cuts taxes and the economy takes off, and Clinton raises taxes and the economy takes off. 

So many things go into a growing or shrinking economy it is impossible to know what one factor will do to the economy. Plausible mechanisms and good theories and quality studies can give us an idea as to what certain policies will do but never certainty. For example, I was a big supporter of the balanced budget amendment during the "contract with America" days and firmly believed that if the deficit was eliminated interest rates would go down. I also remember President Clinton warning that if the deficit went away too soon our prosperity would be threatened. The balanced budget amendment was not passed, but deficits went down anyway. I was surprised when interest rates did not go down, and Clinton was surprised when the economy boomed. It turned out that the budget deficit's effect on interest rates was different than what I thought it would be. Interest rates turned out to be much more complex than I thought.
This complexity is common to many things. Read blogs on the left and right about income inequality and you will find many well thought out and well argued theories about why it has happened, what can be done about it and even whether it’s a bad thing. The theories vary wildly and most are probably wrong. Reading about global warming is similar. Every time I read a good article about it I change my mind. It is just too complex an issue to really know definitively what the truth is. All we have are guesses. Not all guesses are equal of course and the evidence for some guesses are better than others. But if climate researchers can't predict the number of hurricanes from year to year, how can they predict the weather in 50 years with any degree of certainty.Even if a policy produces the expected result, it is impossible to know whether the policy created the result or if it was just luck.  Small sample sizes are notorious for giving false results. I know people who believe a football games outcome is determined by the way they watch the television. If their team scores while they were in the kitchen getting a beer, they watch the rest of the game in the kitchen.

Such irrationality can be easy to fall into in politics as well. Until Carter electing a Democrat president in the 20th century meant getting involved in a war. Republicans are still trying to recover from the fact that the Great Depression happened while a Republican was president. The only way outcomes can really be separated from coincidence is to be tested over and over again. Even policies that have succeeded repeatedly elsewhere are not sure to work. For example, democracy has brought great prosperity to Europe and the US, but the results for Latin America are much more mixed. This is because Latin America is different than Europe and the US. Even in the same country can change drastically over time. The US lost 50,000 people in WW1 but Wilson is viewed by many as a great president. The US has lost 3,000 in Iraq and many of those same people think George W Bush is the worst president ever.

 Because of the complexities of the systems involved and the impossibility of meaningful experimentation it is literally impossible to predict what the correct policies are to address the problems a nation faces. That is why many policies have the exact opposite impact from what was intended. Communism was designed to free the working man and provide prosperity toward the masses; it accomplished the exact opposite of both those goals. World War One made the world safe for Hitler, not democracy. The New Deal prolonged the depression it was meant to end. The list could go on and on.

Given this lack of knowledge, what are the policy implications? If we wait for perfect information, nothing will ever get done and nothing will ever improve. However what we do change, we should change slowly. Governing is like parenting, it is hard to make a positive difference, but easy to screw things up. Changes should be small so if the results are bad they can be undone easily. Incremental changes are to be preferred over radical change. We should beware those who claim certainty and the fanatic who thinks he has all the answers. Modesty should characterize our political debates. When we see a problem we should map out intermediate steps to be taken rather than going to the solution in one leap.

For example, in education many want to completely destroy the public school system on the theory that things in education are so bad they can not get worse. However, if history has taught us anything (and it hasn't) it is that anyone can be killed, and things can always get worse. Thus instead of jumping straight to a complete voucher system, we should take small bites of the apple. First encourage charter schools to experiment in methods. Then open public schools to completion from each other. If these things work then we can move to a complete voucher system where private schools are treated equally with public schools.This may be a frustrating process and take a long time, but the older I get the more wisdom I see for policy makers to adopt a philosophy of "better safe than sorry". 

Share this

Many big changes better than one small change

I agree with most of your post but would emphasize trying multiple approaches simultaneously - which may involve trying each one in a small locality - over making small, incremental changes to the overall society. The latter really slows down the process of discovery. You limit the damage of untried innovation if you test your innovation over a smaller area. For example if you implement a change in half of the US rather than in the whole US, you can get almost as valuable information from that trial as if you were to implement it over the whole of the US, while your risk is reduced. In the meantime you can try a different change in the other half of the US.

I think that monolithic incrementalism is just too slow to get us anywhere. It could be millenia before we significantly explored the possibilities.

For this reason I am to some extent pleased by and hopeful about any political fragmentation that occurs in the world and displeased by unification. As a general policy, I am in favor of states' rights and in favor of secession. There are obvious dangers, of course; but there is also an obvious danger in unification.

Fragmented experimentation also has the advantage of giving us a direct means of comparison. If you attempt an incremental change over the whole society and then you see some other change, you don't even know if the second change was caused by the first or was just a coincidence and was due to a third change. But if you split society up into small areas and try one change in half of the areas and leave the other half alone as a control group, then you can have much more confidence in your ability to interpret the results. Extraneous factors are unlikely to precisely include (or exclude) only the control group, so you have more confidence that visible differences between your groups are related to the change.

An incremental change also has the disadvantage of its being small - perhaps too small for its effects to be noticeable in the midst of the noise. Even increasing the sample size by making the change worldwide may not be enough to distinguish signal from noise, because the noise may be some extraneous factor with worldwide impact, like everybody in the whole world getting affected by some new invention. Bigger changes may be required in order that their effects be felt above the noise. So I would generally recommend more daring changes that each affect smaller numbers of people.

Political fragmentation is key, and the most effective political fragmentation is individual freedom - the ultimate fragmentation. Rather than the state deciding what is wise and good to do and forcing everyone to do this, thereby forcing everyone under their power to participate in a massive single-arm experiment, much more is learned by letting people do whatever they want to do, which creates as many study arms as there are reasonably popular ideas about what to do.

Good idea, but

Federalism is a great idea, both because states can serve as petri dishes for policy experimentation and if things go bad the damage can be contained. Thus the smaller the government the more radical the policy ideas should be. Ideas that work in one place could be tried in others, and if they keep working applied to more and more places. The political trends seems to be going the opposite way, though. The civil rights movement discredited federalism in most people's minds.

Things like this do happen, such as NYC's compstat or Texas's education curriculum. What makes it hard is that each munincipality has its own interest groups to overcome. This means to implement reform you have to fight hundreds of battles. School systems are a good example of this. There are thousands of school systems in the US. However, instead of being hotbeds of experimentation most are held captive by similar coalitions of unions and PTA members. Thus the stagnant situation of the US educational system.


Social "science"?

In science we try to isolate variables to understand complex systems. In politics and policy there is almost no way to do this.

That's a rather amazing claim. There's a vast literature on a wide variety of subjects which are often assumed under the purview of "political science" - voter behavior, gun control, the death penalty... which use a variety of econometric techniques to attempt to make inferences about policy efficiency. Granted, these studies often are open to methodological criticism, but it's not as if real-world systems are necessarily so complex that we should just throw our hands up in frustration.

I personally was a poli sci major until the messiness of it all began to bother me, yes. But I don't begrudge validity to those who dedicate their lives to wading through huge datasets to find confirm statistical significances. There's no need to fall back into irrationally extreme skepticism simply because we can't reproduce social scientific finding in a laboratory, or because of the bogeyman of "complexity". The economy is a complex system, but we can be pretty damn sure that in almost any case raising the demand of a good will raise its price, etc., especially if supply is inelastic. Not all of these conclusions have to be derived through a priori appeals..

We can know things about

We can know things about policy, we just can't know them for sure. The minimum wage debate is a good example. I think I know that raising the minimum wage will cause higher unemployment. However, there are smart people who study economics for a living who would disagree with me. I think they are wrong, but the difficulty of collecting good data, means I can be sure they are wrong.

I think of social science as hopefully approaching truth asymptotically. We should never think of ourselves as having arrived, but always on the journey.