I have argued many, many times that policy mistakes are asymmetric, mostly to no avail, so it's nice to see Justin Wolfers and Betsey Stevenson emphasizing the point that the costs of doing too little are larger than the costs of doing too much, and what this implies for policy:
U.S. Economy Needs Stimulus, Not Soothsayers, By Betsey Stevenson and Justin Wolfers: Here’s something you don’t often hear an economist admit: We have very little idea where the economy will be next year. ...Why? Data are imperfect. Theories are coarse. Models oversimplify. The economy is constantly evolving and can’t be subjected to controlled experiments. Economic cycles are infrequent, so our understanding of them necessarily proceeds very slowly.
None of these drawbacks, though, is fatal to the enterprise. ... Consider the current economic-policy debate. Most forecasters suggest that as the recovery slowly grinds on, unemployment will fall to about 7.5 percent by the end of 2013, from the current 8.3 percent. While this isn’t great progress, it is fast enough that some have argued against further stimulus.
We know, though, that the consensus forecast is highly likely to be wrong. Unemployment could fall to 6.5 percent, or rise to 8.5 percent. Each of these possibilities needs to be considered, and weighed according to its potential benefit or harm.
If unemployment falls to 6.5 percent, there’s no overwhelming reason for concern. ... By contrast, the longer-run consequences could be dreadful, if we find ourselves with 8.5 percent unemployment fully six years after the recession began. ...
In other words, the cost of too little growth far outweighs the cost of too much. If we readily bear the burden of carrying an umbrella when there’s a reasonable chance of getting wet, we should certainly be willing to stimulate the economy when there’s a reasonable risk that doing nothing could yield a jobless generation.