The world’s a complex place and we don’t know a lot about how it works. To make matters worse, it’s not just that we’re ignorant about the world, we’re ignorant about ourselves: We rarely know our own strength. Given our ignorance, how should we interact with the world around us?
William Brainard worked out one answer. With just a little bit of math, he looked at a situation where a policymaker was trying to reach some policy target by pushing on one particular policy lever. Perhaps it’s the central bank trying to reach a GDP target over the next two years by setting short term interest rates, perhaps it’s a doctor trying to reach a blood pressure target by prescribing a medicine, perhaps it’s a dictatorial economic reformer trying to decide how many reforms to enact.
In all these cases, the more the policymaker does, the more uncertain she probably is about the effect of her actions. Experts have a lot of experience raising or lowering interest rates by half a percent so they have a sense of what happens afterward. But they have little recent experience cutting rates by 8 percent in a year in the rich countries. The same is true with blood pressure and economic reforms–experts know a lot about how tinkering works and have a tougher time guessing the effects of bigger changes.
So in many policymaking situations, more action means more uncertainty about the effects of the action. And if your goal is to be close to your target, then more uncertainty is a genuine cost of taking more action.
Hence, Brainard’s Conservatism Principle. Alan Blinder, formerly vice chair of the Fed, summed the Brainard Conservatism Principle this way when he applied it to monetary policy:
Estimate how much you need to tighten or loosen monetary policy to “get it right.” Then do less.
Why do less? Why isn’t the best strategy for hitting the target to try to hit the target? Because more action creates more uncertainty, and you face a real tradeoff.
Extreme versions of this idea turn up in economics and politics from time to time. When the drunk looks for “keys under the lampost” he’s taking the certain, low-risk path. When people say “Don’t just do something, stand there,” they are often noting that people who try to fix problems often create problems of their own. The Brainard Principle shares in those pieces of wisdom but sticks to cases where action is a continuum, where there are a range of options. And it tells us that whatever plain common sense recommends, we probably should do less.
Lessons:
1. If William Tell had been placed further away from his son, he would have aimed higher.
2. It’s prudent to try out medium-size quantitative easing if the Fed doesn’t know much about how QE works in large quantities.
3. When the Hulk shakes hands with people, most folks come away saying, “Gosh, that seemed like a pretty weak handshake.” Also true for the teenage Clark Kent.
Yes, part of the reason to “do less now” when trying out new superpowers like QE is because you can try again later. But even if you only get one shot, Brainard proved that in some cases it’s wise to pull your punches.
Humans, we don’t know our own strength.
Coda: This is my last day here at EconLog, and I want to thank my co-bloggers Bryan and David for helping to make this such a rewarding experience, and I look forward to reading Art’s posts over the coming months. EconLog is a great intellectual environment, I hope it continues for many, many years to come.
I will continue tweeting, so feel to follow: @GarettJones.
READER COMMENTS
Tom West
Apr 30 2013 at 10:33am
Thanks for your posts. I found them interesting and informative.
If you do start your own blog, please ask Brian or David to put up a notice here.
David R. Henderson
Apr 30 2013 at 11:36am
@Garett Jones,
Thanks. I’ve enjoyed having you as a blogging colleague.
B.B.
Apr 30 2013 at 12:45pm
Good points, but it matters what your loss function looks like. Sometimes failure is not an option.
Ike’s strategy at D-Day was not to figure out the optimal and do less. Because failure would be an utter catastrophe, he put everything into it. The invasions might have been successful with less effort, but it was too risky to try that.
Theo Clifford
Apr 30 2013 at 12:48pm
First, it’s a shame to see you depart because your posts at EconLog have been great and I’ve learned a lot. But now I’m going to disagree with you.
The article implicitly assumes that the costs of overshooting are higher than the costs of undershooting. William Tell plays it safe because if he misses, he looks like an idiot – but if he kills his son he looks like an even bigger idiot. Clark Kent would rather seem like a bit of a wimp than betray his secret identity.
It’s not obvious that the central bank should think that overshooting their inflation/NGDP target is worse than undershooting it. But’s pretty obvious that they *do* act in this way.
Obviously the precautionary principle works very differently if you have multiple bites at the apple. It might then be better for the Fed to announce, say, a stronger Evans rule – ‘we will keep printing money until inflation gets too high or unemployment too low.’ That sort of rule means you can combine incremental steps with large action towards concrete goals.
But if you have one shot, and undershooting is as bad or worse than overshooting, you should probably aim at the target.
Greg G
Apr 30 2013 at 1:28pm
I have enjoyed your posts Garett. Thanks and good luck on your future projects.
Various
Apr 30 2013 at 2:24pm
Garett, I should start out by saying that I agree with your conclusion, in most circumstances. However, if the outcome is likely to not be painful, then I think in general it’s better to do “more”, not “less”. This is generally a superior method when one has the liberty of experimentation without significant downside. The results can then be measured and the actions can often be adjusted to create the near-optimal results. This is because interpolation is usually a more accurate method of guessing that extrapolation. To give a crude example, in the U.S. Army artillery, if the first shot is short, the folks in fire control intentionally make the second shot long. They can then interpolate the results and be on target with shot number three. Being short with both the first and second shots is a no-no. Of course, exceptions are made to this convention if friendly forces are on the “long” side (i.e., nasty downside consequences if one is wrong).
My father (who was in the artillery, and later earned a PhD in Chemistry from the University of Illinois) told me a funny story that occurred during a training exercise. A General had made a surprise visit to the battery and asked the lieutenant in charge to fire on a certain target. The lieutenant, who my father claimed was incompetent, was “short” of the target seven straight times. The lieutenant was so nervous he attempted to light a cigarette and inserted it between his lips backwards.
E. Barandiaran
Apr 30 2013 at 2:39pm
Garett, thanks for your interesting posts.
Your last one is about economic policy and frankly I was surprised by your reference to W. Brainard. In his old, well-known paper Brainard makes reference to J. Tinbergen and H. Theil and he could have referred to B. Hansen and R. Frisch, all pioneers of the theory of economic policy. Actually, Tinbergen and Frisch were rewarded with the first Nobel Prize largely because of their contribution to economic policy. All that was before the Rational Expectations revolution and the Lucas Critique.
The old theory has survived in public debates of economic policy, particularly monetary and fiscal policies. A new one taking into account the critique to the old theory has recently been circulated by N. Acocella and G. Di Bartolomeo in collaboration with their mentor A. Hughes Hallett, now with GMU’s School of Public Policy; see their new book
http://www.amazon.com/Theory-Economic-Policy-Strategic-Context/dp/1107023866/ref=tmm_hrd_title_0?ie=UTF8&qid=1367337335&sr=1-1
I’ve yet to finish the book but it seems the authors have made important contributions that indeed amount to a new theory. They go far beyond of what Brainard argued about monetary policy under uncertainty and about your immediate concern in your post. But after completing my Ph. Thesis in 1970 (I relied on the old theory and used the Markowitz-Tobin-Brainard view of monetary policy to search for an instrument), I have been deploring the use of the old theory in most public discussions of economic policy for failing to deal properly with uncertainty, or more precisely with the many sources of uncertainty. If you’re familiar with the new theory, I’d like to know your views on it.
Grieve Chelwa
Apr 30 2013 at 4:15pm
Thanks Garett for all the wonderful and informative posts. I learnt a lot from them.
Philo
Apr 30 2013 at 4:57pm
I would phrase Theo Clifford’s point like this: “If you have one shot, and undershooting is exactly as bad as overshooting, you should simply aim at the target, regardless of the amount of uncertainty as to whether you will hit it.” If undershooting were worse, you should aim high—-the worse, the higher. If overshooting were worse, you should aim low—-the worse, the lower. In the first situation, though you would prefer to be certain of hitting your target, the (unfortunate) uncertainty is irrelevant to your choice of action.
Ken P
Apr 30 2013 at 10:55pm
I’ve enjoyed your posts, Garett! Good luck in the future!
Aaron Zierman
Apr 30 2013 at 11:15pm
I wonder if “aiming low” has another effect. If we didn’t hit our target, but were knowingly aiming low, was our “ammo” wrong? Or did we just not use enough? For example:
Say target is 5% growth, 2% inflation.
We want to take action of ensuring low interest rates, and “quantitative easing”. Call it “X”. Based on this blog, instead of doing full “X”, we instead do 1/2 X.
We miss target. Instead we get 2.2% growth, 2.1% inflation. Hmmm…what was the problem?
So we do another 1/2 X.
Slight increase, but still far shy of target.
Let’s boost it to 3/4 X.
No change. So now we’ve done 1.75 of our original “X” and still have not achieved our goal.
So my question is: does this policy have a negative effect in that we can never have a straightforward answer on whether our approach is simply incorrect, or if the force behind it is insufficient?
Thomas Colthurst
May 28 2013 at 11:08am
Computer scientists have also investigated this problem and come to some different conclusions.
Most specifically: if you are playing an iterated (rather than a one-shot) game, then it is worthwhile to sometimes pick the less tried (more uncertain) options, because sometimes you will discover that these options are superior.
In the lingo of the multi-armed bandit field, sometimes you want to “explore” (as described in the previous paragraph) and sometimes you want to “exploit” (pick the option which your current knowledge says is best).
The simple strategy of exploring some fixed percentage of the time works surprisingly well in practice; more sophisticated strategies start out by exploring a lot and end up exploiting 100% of the time in the limit.
Comments are closed.