Monday, March 19, 2018

Rules and Meta-Rules

Really interesting Econlib post here.

Basically, one economist says (paraphrasing, not actually quoting here), "I must be irrational, because I could switch to generic diet soda, which is cheaper and likely just as good. But I don't bother, because I'm stuck in some bad default mode. Even if it's not just as good, it's probably at least worth trying."

Another economist says, Hold on there! The brain power necessary for decision-making is a truly scarce resource. You aren't being irrational by creating some simple rules that economize on this resource, such as "always buy Diet Coke." (He says a lot more, but this is my blog so I'm summarizing my own take-away.)

This is an important and deep point. You can't simply re-analyze every single decision from scratch. You need some simple rules, such as "Leave work at 4:50", "Set alarm clock for 6:30", and "If we're low on our household stock of Diet Coke, pick up more at the store." You also need a set of meta-rules for resetting these rules of thumb. "If the boss complains about my punctuality twice, set alarm clock for 6:20 instead of 6:30." Or "If the expected benefit of changing from Coke to generic brand exceeds X, switch." The problem here is that you need a rule to even alert you to the possibility of changing a decision-rule. Maybe the benefit of switching from Coke to generic is enormous, so big that almost anyone would agree that it's irrational not to attempt the switch. But you don't just get to know that. You have to sit down for a moment, jot down some figures, grind out a simple calculation, and see the answer. Someone who did this all the time for every little decision would be almost paralyzed by their constant cost-benefit checking. (Flip over a used envelope, or grab a sticky note? Pen or pencil? Or just use Excel? GHAA!) So your meta-rules require not just a set of rules about thresholds that overturn existing rules, but for when you will even bother to check whether the threshold has been crossed.

This is a theme I've discussed before. I think the whole economic irrationality/behavioral economics thing is way overblown and overdone. Examples of people allegedly behaving irrationally usually have an explanation that is fully compatible with a "rational actor" model. Just add in, say, limited computing power or limited information and you get back to "imperfect but serviceable decision rules."  Of course some clever person can pick these rules apart and find that they are occasionally wrong or even that they tend to be biased wrong. Such rules can still be useful.

1 comment:

  1. > The problem here is that you need a rule to even alert you to the possibility of changing a decision-rule.

    one part of that rule should be that you allow anyone in the public to alert you (yes, *anyone* – that's clearly superior than filtering by prestige or whatever, *if* you can find a way to organize that so it's cheap and manageable).

    see:

    http://fallibleideas.com/paths-forward

    http://curi.us/2068-my-paths-forward-policy

    ReplyDelete